New data provided to Search Engine Journal shows that the sites Google is ranking in AI Overviews varies by time and industry, offering an explanation of volatility in AIO rankings. The new research shows what industries are most impacted and may provide a clue as to way.
AIO Presence Varies Over Time and By Industry.
The research was provided by BrightEdge using their proprietary BrightEdge Generative Parser technology that tracks AI Overviews, detects patterns and offers insights useful for SEO and marketing.
Healthcare, Education, and B2B Technology topics continue to show greater presence in Google’s AI Overviews. Healthcare and Education are the two industries where BrightEdge saw the strongest growth as well as stability of which sites are shown.
Healthcare has the highest AIO presence at 84% as of late February 2025. AIOs shown for Education topics show a consistent growth pattern, now at 71% in February 2025.
The travel, restaurant and insurance sectors are also trending upward, with the travel queries being a notable trend. Travel had zero AIO presence in May 2024 but that’s completely different now. Travel is now up to 20-30% presence in the AIO search results.
The presence of restaurant related topics in AIO are up from 0 to 5%, suggesting a rising trend. Meanwhile insurance queries have grown from 18% of queries in May 2024 to a whopping 47% of queries by February 2025.
B2B technology queries that trigger AIO are at 57%. These kinds of queries are important because they are typically represent research related by people involved in decision making. Purchase decisions are different than with consumer queries. So the fact that 57% of queries are triggering AIOs may be a reflection of the complexity of the decision making process and the queries involved with that process.
Let’s face it, technology is complex and the people using it aren’t expert in concepts like “data modeling” and that’s the kind of queries BrightEdge is seeing, which could be reflective of the end user wrapping their minds around what the technology does and how it benefits users.
Having worked with B2B technology it’s not unusual for SaaS providers to use mind numbing jargon to sell their products but the decision makers or even the users of that technology aren’t necessarily going to understand that kind of language. That’s why Google shows AI Overviews for a keyword phrase like associative analytics engine instead of showing someone’s product.
Finance related queries, which had been on a moderate growth trend have doubled from 5% of queries in May 2024 to 10% of queries in February 2025.
Here’s the takeaway provided by BrightEdge:
B2B Tech is at 57%, in Feb-25. Finance has been growing moderately and doubled from 5% in May-24 to 10% in Fed-25
Ecommerce 4% (down from 23% in May-24). Entertainment has dropped to 3%.
Ecommerce and Entertainment presence drops from suggests more testing and alignment with traditional Google search where users can engage in platform experiences. For Ecommerce, the use of features like product grids may be the reason. Traditional search provides more in-platform experiences.
What Does This Mean?
This volatility could reflect variable quality of complex user queries. Given that these are complex queries that are triggering AIO then it may be reasonable to assume that they are longtail in nature. Longtail doesn’t mean that they’re long and complex queries, they can also be short queries like “what is docker compose?”
Screenshots of Google trends shows that more people query Docker Compose than they do What is Docker Compose or What is Docker. Why do more people do that?
Screenshot Of Google Trends
It’s clearly because people are querying Docker Compose as a navigational query. And you can prove that Docker Compose is a navigational query because Google’s search results don’t show an AIO for the query “Docker Compose” but it does show AIO for the other two.
Screenshot Shows SERPs For Docker Compose
Screenshot Shows “What Is” Query Triggers AIO
Changes In AIO Patterns: Gains For Authoritativeness
An interesting trend is that queries for some topics correlated to answers from big brand sites. This is interesting because it somewhat mirrors what happened with Google’s Medic update where SEOs noticed that non-scientific websites no longer ranked for medical queries. Some misunderstood this as Google betraying a bias for big brand sites but that’s not what happened.
What happened in that update was not limited to health related topics. It was a widespread effect that was more like a rebalancing of queries to user expectations- which means this was all about relevance. A query about diabetes should surface scientific data not herbal remedies.
What’s happening today with AIO, particularly with AIO, is a similar thing. Google is tightening up the kind of content AIO is showing to users for medical and technology queries.
Is it favoring brands or authoritativeness? The view that Google has favored brands is shallow and lacks substance. Google has consistently shown a preference for ranking what users expect to see and there are patents that support that observation. SEOs who expect to see rankings based on their made for search engines links, optimized for search engines content, and naïve “EEAT optimized” content completely miss the point of what’s really going on in today’s search engines that rank content based on topicality, user preferences and user expectations. Trustworthy signals of authoritativeness very likely derive from users themselves.
Here’s what BrightEdge shared:
“For example, in the healthcare category, where accuracy and trustworthiness are paramount, Google is increasingly showing search results from just a handful of websites.
Content from authoritative medical research centers account for 72% of AI Overview answers, which is an increase from 54% of all queries at the start of January.
15-22% of B2B technology search queries are derived from the top five technology companies, such as Amazon, IBM, and Microsoft.”
Takeaways:
AIO Presence Varies by Industry and Time
There is growth in AIO visibility for Healthcare, Travel, Insurance, and B2B Technology
Declining presence of AIO in Ecommerce and Entertainment
AIO patterns indicate a preference for authoritative sources. AIO results are increasingly sourced from authoritative sites, particularly in Healthcare and B2B Tech. In B2B Tech, 15-22% of AIO responses come from the top five companies. This shift may mirror previous Google updates like the Medic Update that appeared to rebalance search results based on authoritativeness and user expectations.
In 2025, the extent to which you adapt to emerging technologies, changing user expectations, and evolving search engine algorithms will determine if you’ll thrive or struggle to stay relevant.
Staying ahead of emerging trends is essential for maintaining a fast, secure, and user-friendly website.
Optimizing performance, strengthening security measures, and enhancing user experience will be key factors in staying competitive.
Artificial intelligence has transformed how websites interact with visitors, making online experiences more personalized, engaging, and efficient.
Use AI For Higher Conversion Rates
AI-driven personalization allows websites to deliver tailored content and product recommendations based on user behavior, preferences, and past interactions to create an intuitive experience.
The result? Visitors remain engaged, increasing conversions.
Chatbots and AI-powered customer support are also becoming essential for websites looking to provide instant, 24/7 assistance.
These tools answer common questions, guide users through a website, and even process transactions, reducing the need for human intervention while improving response times.
And they’re gaining in popularity.
71% of businesses in a recent survey either already have a chatbot integrated into their sites and customer service processes or plan to get one in the near future.
And they’re reaping the benefits of this technology; 24% of businesses with a chatbot already installed report excellent ROI.
Use AI For Speeding Up Website Implementation
AI is also revolutionizing content creation and website design.
Based on user data, automated tools can generate blog posts, optimize layouts, and suggest design improvements.
This streamlines website management, making it easier for you to maintain a professional and visually appealing online presence.
For example, many hosting providers now include AI-powered website builders, offering tools that assist with design and customization. These features, such as responsive templates and automated suggestions, can make building and optimizing a website more efficient.
2. Voice Search & Conversational Interfaces
Voice search is becoming a major factor in how users interact with the web, with more people relying on smart speakers, mobile assistants, and voice-activated search to find information.
To put this into perspective, ChatGPT from OpenAI reportedly holds 60% of the generative AI market, performing more than one billion searches daily. If just 1% of those are via its voice search, that equates to 10 million voice searches every day on ChatGPT alone.
Reports estimate 20.5% of people globally use voice search daily. And these numbers are increasing.
You need to adapt by optimizing for conversational SEO and natural language queries, which tend to be longer and more specific, making long-tail keywords and question-based content more important than ever.
To stay ahead, websites should structure content in a way that mimics natural conversation:
Google continues to refine its ranking algorithms, with Core Web Vitals playing a critical role in determining search visibility.
Implement Core Web Vital Data & Monitor Website Speed
These performance metrics, Largest Contentful Paint (LCP), First Input Delay (FID), and Cumulative Layout Shift (CLS), measure how quickly a page loads, how responsive it is, and how stable its layout appears to users.
Websites that meet these benchmarks not only rank higher in search results but also provide a better overall user experience.
One study found that pages ranking in the top spots in the SERPs were 10% more likely to pass CWV scores than URLs in position 9.
Ensure Your Website Is Faster Than Your Competitors To Rank Higher
As part of the prioritization of performance, mobile-first approach remains essential; Google prioritizes sites that are fast and responsive on smartphones and tablets.
Ensuring faster load times through optimized images, efficient coding, and proper caching techniques can make a significant impact on search rankings.
Leverage Structured Data To Tell Google What Your Website Is About
Structured data, on the other hand, helps search engines better understand a website’s content, improving the chances of appearing in rich snippets and voice search results.
4. Mobile-First & Adaptive Design
With mobile devices accounting for the majority of web traffic, mobile optimization remains a top priority in 2025.
Google’s mobile-first indexing means that search engines primarily evaluate the mobile version of a site when determining rankings.
A website that isn’t optimized for mobile results in overall poor performance, lower search rankings, and a frustrating user experience.
To keep up, many websites are adopting:
Adaptive design – Ensures that websites adjust dynamically to different screen sizes, providing an optimal layout on any device.
Progressive Web Apps (PWAs) – Combine the best features of websites and mobile apps, offering faster load times, offline capabilities, and app-like functionality without requiring a download.
Best practices for a seamless mobile experience include responsive design, fast-loading pages, and touch-friendly navigation.
Optimizing images, minimizing pop-ups, and using mobile-friendly fonts and buttons can also greatly enhance usability.
5. Enhanced Website Security & Data Privacy
Cyber threats are becoming more sophisticated.
You must take proactive measures to protect your websites from attacks, data breaches, and unauthorized access.
Implementing strong security protocols not only safeguards sensitive information but also builds trust with visitors.
Key security measures include:
SSL certificates – Encrypt data transmitted between users and a website, ensuring secure connections—something that search engines and users now expect as a standard feature.
Multi-Factor Authentication (MFA) – Adds an extra layer of security by requiring multiple verification steps before granting access, reducing the risk of compromised credentials.
Zero-trust security models – Ensures that all access requests, even from within a network, are continuously verified, minimizing potential security gaps.
Beyond technical defenses, compliance with evolving privacy laws such as GDPR and CCPA is essential.
You must be transparent about how they collect, store, and process user data, providing clear consent options and maintaining privacy policies that align with current regulations.
6. Sustainability & Green Web Hosting
Every website, server, and data center requires energy to function, contributing to global carbon emissions.
Optimizing websites through lighter code, efficient caching, and reduced server load also plays a role in minimizing environmental impact.
Choosing a hosting provider that values sustainability is an important step toward a greener web.
AI tools can assist in creating blog posts, product descriptions, and videos with minimal manual input, helping businesses maintain a steady content flow efficiently.
Beyond static content, interactive features like quizzes, calculators, and AR are becoming key for user engagement.
These elements encourage participation, increasing time on site and improving conversions.
To integrate interactive features smoothly, a hosting provider that supports interactive plugins and flexible tools can help keep websites engaging and competitive.
8. The Role of Blockchain in Web Security
Blockchain is emerging as a tool for web hosting and cybersecurity, enhancing data security, decentralization, and content authenticity.
Unlike traditional hosting, decentralized networks distribute website data across multiple nodes, reducing risks like downtime, censorship, and cyberattacks. Blockchain-powered domains also add security by making ownership harder to manipulate.
Beyond hosting, blockchain improves data verification by storing information in a tamper-proof ledger, benefiting ecommerce, digital identity verification, and intellectual property protection.
9. The Importance of Reliable Web Hosting
No matter how advanced a website is, it’s only as strong as the hosting infrastructure behind it. In 2025, website performance and uptime will remain critical factors for success, impacting everything from user experience to search engine rankings and business revenue.
Scalable hosting solutions play a crucial role in handling traffic spikes, ensuring that websites remain accessible during high-demand periods.
Whether it’s an ecommerce store experiencing a surge in holiday traffic or a viral blog post drawing in thousands of visitors, having a hosting plan that adapts to these changes is essential.
Reliable hosting providers help mitigate these challenges by offering scalable infrastructure, 100% SLA uptime guarantees, and built-in performance optimizations to keep websites running smoothly.
Features like VPS and dedicated hosting provide additional resources for growing businesses, ensuring that increased traffic doesn’t compromise speed or stability. Investing in a hosting solution that prioritizes reliability and scalability helps safeguard a website’s long-term success.
Future-Proof Your Website Today
The digital landscape is changing fast, and staying ahead is essential to staying competitive.
From AI-driven personalization to enhanced security and sustainable hosting, adapting to new trends ensures your site remains fast, secure, and engaging. Investing in performance and user experience isn’t optional, it’s the key to long-term success.
Whether launching a new site or optimizing an existing one, the right hosting provider makes all the difference.
Bluehost offers reliable, high-performance hosting with built-in security, scalability, and guaranteed uptime, so your website is ready for the future.
Get started today and build a website designed to thrive.
OpenAI published a new write-up about elevated errors in ChatGPT that significantly increased failed conversation attempts. The issue was caused by a misconfigured internal experiment.
According to OpenAI:
“On February 19, 2025, from 9:48 AM to 11:19 AM PT, ChatGPT experienced a service degradation, leading to a significant increase in failed conversation attempts. This resulted in blank responses for many users.
The root cause was a misconfigured internal experiment that unintentionally triggered a surge in traffic, overwhelming our inference infrastructure. This increase in load led to saturation of compute resources, causing failures in generating responses.
After identifying the root cause, we took immediate action by temporarily shedding load from free-tier users to stabilize the system. As capacity was restored, paid users gradually recovered, and the full service was restored by 11:19 AM PT.”
OpenAI Continues To Work On Solutions
The incident response goes on to note that they continue to work on changes that will prevent similar outages from happening, writing:
“Stronger Safeguards: Building better protections around experiment changes and configurations by moving from a uniform approval process to a risk-based model to ensure safer rollouts of experiments.
Faster Root Cause Identification: Automating notifications for relevant changes and experiments to more quickly identify root causes of increased failures.”
A recent analysis by xfunnel.ai examines citation patterns across major AI search engines.
The findings provide new insight into how these tools reference web content in their responses.
Here are the must-know highlights from the report.
Citation Frequency Differs By Platform
Researchers submitted questions across different buyer journey stages and tracked how the AI platforms responded.
The study analyzed 40,000 responses containing 250,000 citations and found differences in citation frequency:
Perplexity: 6.61 citations per response
Google Gemini: 6.1 citations per response
ChatGPT: 2.62 citations per response
ChatGPT was tested in its standard mode, not with explicitly activated search features, which may explain its lower citation count.
Third-Party Content Leads Citation Types
The research categorized citations into four groups:
Owned (company domains)
Competitor domains
Earned (third-party/affiliate sites)
UGC (user-generated content)
Across all platforms, earned content represents the largest percentage of citations, with UGC showing increasing representation.
Affiliate sites and independent blogs hold weight in AI-generated responses as well.
Citations Change Throughout Customer Journey
The data shows differences in citation patterns based on query types:
During the problem exploration and education stages, there is a higher percentage of citations from third-party editorial content.
UGC citations from review sites and forums increase in the comparison stages.
In the final research and evaluation phase, citations tend to come directly from brand websites and competitors.
Source Quality Distribution
When examining the quality distribution of cited sources, the data showed:
High-quality sources: ~31.5% of citations
Upper-mid quality sources: ~15.3% of citations
Mid-quality sources: ~26.3% of citations
Lower-mid quality sources: ~22.1% of citations
Low-quality sources: ~4.8% of citations
This indicates AI search engines prefer higher-quality sources but regularly cite content from middle-tier sources.
Platform-Specific UGC Preferences
Each AI search engine shows preferences for different UGC sources:
Perplexity: Favors YouTube and PeerSpot
Google Gemini: Frequently cites Medium, Reddit, and YouTube
ChatGPT: Often references LinkedIn, G2, and Gartner Peer Reviews
The Third-Party Citation Opportunity
The data exposes a key area that many SEO professionals might be overlooking.
While the industry often focuses on technical changes to owned content for AI search optimization, this research suggests a different approach may be more effective.
Since earned media (content from third parties) is the biggest citation source on AI search platforms, it’s important to focus on:
Building relationships with industry publications
Creating content that others want to cover
Contributing guest articles to trusted websites
Developing strategies for the user-generated content (UGC) platforms that each AI engine prefers
This is a return to basics: create valuable content that others will want to reference instead of just modifying existing content for AI.
Why This Matters
As AI search is more widely used, understanding these citation patterns can help you stay visible.
The findings show the need to use different content strategies across various platforms.
However, maintaining quality and authority is essential. So don’t neglect SEO fundamentals in pursuit of broader content distribution.
Top Takeaway
Invest in a mix of owned content, third-party coverage, and presence on relevant UGC platforms to increase the likelihood of your content being cited by AI search engines.
The data suggests that earning mentions on trusted third-party sites may be even more valuable than optimizing your domain content.
Recent studies by Gartner and Adobe show that generative AI is becoming a key tool in marketing.
Almost three-quarters of marketing teams now use GenAI, and most consumers are comfortable with AI in advertising.
AI Adoption In Marketing
A survey by Gartner of 418 marketing leaders found that 73% of marketing teams use generative AI.
However, 27% of CMOs say their organizations have limited or no use of GenAI in their marketing campaigns.
Correlation With Top Performers
Marketing teams that consistently exceed targets and meet customer acquisition goals are adopting AI faster than competitors.
Greg Carlucci, Senior Director Analyst in the Gartner Marketing Practice, states:
“The most successful marketing organizations are leading the way when it comes to GenAI adoption.”
Most marketers are using GenAI for:
Creative development (77%)
Strategy work (48%)
Campaign evaluation (47% reporting benefits)
Challenges With Generative AI
Despite spending almost half their budgets on campaigns, 87% of CMOs faced performance problems last year, and nearly half had to end underperforming campaigns early.
The Gartner study found:
“On average, 87% of CMOs report they experienced campaign performance issues in the last 12 months, with 45% reporting that they sometimes, often, or always had occasion to terminate campaigns early in the last year due to poor performance.”
CMOs identified several departments as barriers to their success:
Finance (31%)
Executive leadership (26%)
Sales (26%)
Opportunities With Generative AI
Adobe’s research highlights personalization as the primary AI opportunity for marketers.
Heather Freeland, Chief Brand Officer at Adobe, notes:
“Across all industries, there is an insatiable demand for content as customers expect every encounter with a brand to be personalized.”
She adds:
“Just when this challenge seemed insurmountable, the emergence of generative AI is presenting creative and marketing teams with a new way to keep pace with customer demands while also breaking through with their brands.”
The study finds that 97% of marketers believe mass personalization is achievable with AI, but most find it challenging without appropriate tools.
AI Acceptance Among Consumers
Consumers say that knowing content was created by AI either makes them more engaged or does not change their engagement at all.
Adobe’s study found:
Three in four consumers surveyed agree that knowing content was AI-produced would either improve or not impact their likelihood of engaging with it.
Consumers are even willing to share their data for a better AI-driven experience.
Adobe’s study finds the top data points consumers are willing to share include:
“… past purchases (56%), products they’ve viewed (52%), their gender (47%), age (41%), and language (35%).”
Generational Differences
Different age groups prefer personalization in different channels.
According to Adobe’s research:
“Gen Z respondents show a higher affinity for personalized content from the consumer electronics industry, particularly music (45%) and video games (43%)…
This contrasts with Baby Boomers, who prefer personalization in retail industry content, specifically from grocery stores (46%).”
The study also found:
“Millennials prefer personalized email campaigns (45%) and website content (40%), while Gen Z values social media personalization (51%).”
Measurable Results
Adobe reports that the implementation of GenAI tools delivered performance improvements.
Its report states:
“… in one of our first generative AI-powered email tests, we used the tool to quickly build and test five versions of an Adobe Photoshop email. It delivered a more than 10% increase in click-through rates, and a subsequent test reported a 57% increase in click rates for an Adobe Illustrator email.”
Additionally:
“Testing scale and speed transformed our approach to content optimization, significantly enhancing our marketing performance and efficiency.”
What This Means
Generative AI is shifting from a novel technology to a standard practice within marketing.
Marketing departments are facing tighter budgets while consumer demand for personalized content grows. Generative AI offers a potential solution to create personalized content at scale.
Further, using AI to personalize marketing messages will unlikely impact consumer perception of your brand. Some marketers believe it may even improve retention.
Adobe’s research suggests:
“Over one in four (26%) marketer respondents agree that AI-powered personalization will increase consumer brand loyalty.”
If you want to incorporate AI into your advertising strategy but are unsure where to start, data suggests that the best approach is to enhance personalization.
While AI chatbot traffic currently represents a tiny percentage of overall traffic, the data shows early evidence for the value of citations and mentions.
AI chatbot adoption is skyrocketing, referral traffic to websites is growing, and traffic quality is high.
Adoption
ChatGPT has over 400 million weekly users as of January 2025.1
Semrush, 12/24: Most ChatGPT users are from the U.S. (25%) or India (12%), followed by India, Brazil, the UK and Germany. 70% are male, and over 50% are between 18 and 34 years old.
Higher Visibility, 02/25: 71.5% of consumers use ChatGPT for searching but complementary to Google, not as a replacement.
Ahrefs, 02/25: 63% of websites receive at least some traffic from AI sources. Only 0.17% of total visits came from AI Chatbots, with top sites achieving up to 6%.
98% of AI traffic comes from three AI chatbots: ChatGPT (> 50%), Perplexity (30.7%), and Gemini (17.6%).
Smaller sites get proportionally more visits from AI.
Semrush, 02/25: The generative AI market was valued at $67 billion in 2024 and is expected to grow annually by 24.4% through 2030.
Semrush, 12/24: ChatGPT referrals to websites grew by 60% between June and October.
Semrush, 02/25: ChatGPT’s reach has expanded dramatically, sending traffic to over 30,000 unique domains daily in November 2024, up from less than 10,000 in July.
Online services, education, and mass media are getting the most referral traffic from ChatGPT after filtering out authentication URLs. Retail, finance, and healthcare show lower volumes.
Growth Memo, 02/25: The quality of AI chatbot traffic is superior in several key metrics:
The average session duration is 10.4 minutes for AI chatbot referrals versus 8.1 minutes for Google traffic.
Users view more pages: 12.4 pages on average for AI chatbot referrals compared to 11.8 for Google traffic.
Impact On Purchase Decisions:
Adobe, 10/24: 25% of Britons use AI while shopping online.
AI usage rose 10x between July and September to 10 billion visits to UK retail websites and ~100 million products.
Most shoppers are looking for deals:
In an Adobe survey of 5,000 U.S. consumers, 7 in 10 respondents who have used generative AI for shopping believe it enhances their experience. Additionally, 20% of respondents turn to generative AI to find the best deals, followed by quickly finding specific items online (19%) and getting brand recommendations (15%).
Semrush, 02/25: 46% of ChatGPT queries use the Search feature.
The research paper “A comparative study on the effect of ChatGPT recommendation and AI recommender systems on the formation of a consideration set” by Chang et al. looked at 471 consumers to understand:
Whether ChatGPT impacts consumer choices.
The process that impacts choices.
The impact on products with low-brand awareness vs. high-brand awareness.
Results:
ChatGPT does influence the consumer purchase journey and products recommended by ChatGPT are more likely to be adopted.
Products with low brand awareness see higher trust after a recommendation from ChatGPT.
My Take
ChatGPT had 560 million unique worldwide visitors in December 2024, compared to Google’s 6.5 billion. For comparison, that’s still small but about the size of X/Twitter today.
ChatGPT sending more referral traffic to a diverse list of domains is probably a strategic move to win the web over and establish itself more as an alternative to Google. I don’t think OpenAI has to do that. I think they strategically chose to.
So far, it seems young men in the U.S., BRIC, and Europe are the major users of ChatGPT. If that’s your target audience, optimizing for AI chatbot visibility should be a higher priority.
To be crystal clear, I don’t think anybody has to optimize for AI chatbot visibility. I’m confident that most industries will be fine doing classic SEO for years to come. Some will even be fine in a decade. However, you can’t unsee the rapid adoption, which leads us to a situation where two things are true: classic SEO still works, and there is a first-mover advantage on AI chatbots.
How Can You Grow Your AI Chatbot Visibility?
Improving AI chatbot visibility is a mix of known and new levers.
Crawlability
Being visible on AI chatbots starts with being visible to their crawlers. Crystal Carter, head of SEO Commus at Wix, calls this “retrievability.”
Groomed XML sitemaps, strong internal linking, fast server response, and clean HTML are a good start.
LLM crawlers are less forgiving than Google when it comes to JavaScript and client-side rendering for critical SEO components. Avoid at all cost!
Brand Strength
Ziff Davis, 11/24: A Ziff Davis study compares Domain Authority in curated (OpenWebText, OpenWebText2) with uncurated public web indices (Common Crawl, C4) to investigate how major AI companies like OpenAI, Google, and Meta trained their large language models. The unsurprising conclusion is that AI developers prefer curated text to train their models, naturally giving commercial publishers more visibility.
Semrush, 12/24: Google tends to show larger domains, ChatGPT smaller ones. The opposite is true for transactional searches: Search GPT prefers larger domains, Google smaller ones.
Seer, 01/25: Backlinks showed no correlation with AI chatbot visibility.
Organic Ranks
Seer, 01/25: Brands ranking on page 1 of Google showed a strong correlation (~0.65) with LLM mentions. Bing rankings also mattered, but a little less (~0.5–0.6).
Semrush, 02/25: The overlap between Google, Perplexity, and ChatGPT search is low (25-35% on average). However, the overlap between ChatGPT search and Bing is much higher (average = 7 domains) than with Google (4 domains).
Semrush, 02/25: YouTube is the third largest domain by referral traffic from ChatGPT. Facebook, LinkedIn, and GitHub are in the top 10.
Growth Memo, 02/25: Amazon, eBay, and Walmart dominate in Google Search just as much as in AI chatbots.
My Take
There is a big question of how important backlinks are for AI chatbot visibility. I think there is a trap to think they have a direct impact. The way I understand the data is that they help with Google/Bing visibility, which passively translates to AI chatbot visibility. They might also help with LLM crawler discoverability. So, they’re still important but not as much as the content itself.
The biggest lever seems to be citable content on and off of Google: Industry reports with exclusive research and data, original surveys and case studies, and thought leadership content from recognized experts.
I wouldn’t restrict myself from optimizing for AI chatbot visibility as a small business with little to no visibility on classic search engines.
Ecommerce is an outlier because the journey is so much more transactional than for B2B or media. On one hand, the strong visibility of big ecommerce platforms like Amazon provides a direct path for AI chatbot visibility for merchants. On the other hand, integrating with programs like Perplexity’s Buy With Pro seems worth trying out.
How Are People Searching On AI Chatbots?
Consumers use AI chatbots differently than Google unless they turn on search features.
Semrush, 02/25: 70% of ChatGPT queries represent entirely new types of intent that don’t fit traditional search categories (navigational, informational, commercial, transactional).
Users are asking longer, more complex questions, with non-search-enabled ChatGPT prompts averaging 23 words compared to 4.2 words when search is enabled.
Higher Visibility, 02/25: People use different AI chatbots for different user intents, e.g., Google for initial product research, ChatGPT for product comparison, and Instagram for discovering new products. However, almost 80% stick to traditional search engines for informational searches.
Growth Memo, 02/25: AI chatbots send significantly more traffic to homepages (22% on average) compared to Google (10%) yet still maintain higher engagement metrics. This trend suggests that AI chatbots are effectively preparing users for brand interactions.
My Take
It’s fascinating to see that when people turn on Search in ChatGPT, they use shorter queries and emulate their behavior on Google. I wonder if this behavior sticks over the long term or not. If so, we can assume a stronger carryover from players who dominate classic search engines today to AI chatbots. If not, it might open the field to new players.
I’ve long been dissatisfied with our broad classification of user intents (information, navigational, etc.). We had this wrong for a long time. It’s too coarse. 70% of use cases are likely task-related and don’t fit our model for classic search engines. AI chatbots are more than search engines but solve the same problems, just with different means. That’s also where I see Google lagging behind: Consumers already associate AI chatbots with tasks rather than finding information.
What Challenges Are Associated With AI Chatbots?
AI chatbots make for a compelling marketing channel but put marketers in front of tracking and bias problems.
We can track the referral source for almost all AI chatbots, but some traffic can still fall into the direct traffic bucket.
Citations in ChatGPT typically include a “utm_source=chatgpt.com” parameter, but links in search results don’t have the parameter.2
Ahrefs, 02/25: AI traffic is likely underreported because AI chatbots like Copilot get clustered into direct while they’re actually referrals.
Brand Bias
Semrush, 12/24: Consumers and users are skeptical about AI output. 50% say they trust it more when it’s been reviewed by a human.
In the paper “Global is Good, Local is Bad?” Kamruzzaman et al. conducted experiments with fill-in-the-blank questions across four product categories and 15 countries (English only). The researchers studied the effect of:
Brand attribute bias: global vs. local brands.
Socio-economic bias: luxury vs non-luxury brands.
Geo bias: local brands when the domestic country is specified.
Results:
LLMs across multiple models (GPT-4o, Llama-3, Gemma-7B, Mistral-7B) consistently associate global brands with positive and local brands with negative attributes.
LLMs tend to recommend luxury brands to people from high-income countries. In contrast, non-luxury brands are more commonly suggested for people from low-income countries, even when models were given the flexibility to suggest the same brands for both groups.
The underlying reasons are that local brand names are underrepresented in LLM training data, and large companies can afford larger marketing campaigns and, therefore, create more bias.
In the paper “Generative AI Search Engines as Arbiters of Public Knowledge: An Audit of Bias and Authority” by Li et al., researchers tested how ChatGPT, Bing Chat, and Perplexity answer questions about four major topics: climate change, vaccination, alternative energy, and trust in media. They wanted to see if the AI showed bias in its answers and how it tried to appear trustworthy.
The results:
The AI tends to match the emotion of the question. If you ask a negative question, you get a negative answer.
Different topics got different emotional treatment, e.g., vaccination and alternative energy got more positive responses than climate change and media trust.
Bing Chat and Perplexity heavily cite news media and businesses.
Heavy reliance on U.S. sources (65% of sources), even when used in other countries.
Too many commercial/business sources, especially for topics like alternative energy.
Some models mix unreliable sources with good ones.
Answers often include uncertain language and hedging to avoid taking strong positions.
My Take
We’re used to significant tracking gaps from Google and Bing, so unless AI chatbots try to persuade site owners with more data, we’ll have to continue to operate with aggregate data, as I mentioned in Death of the Keyword.
AI chatbot bias is serious. User trust is key to winning, so I assume AI developers are aware and try to solve the problem. Until then, we have to factor bias in with our optimization strategies and do our best to clearly indicate the target audience for our product in our content.
Conclusion: Where It’s All Going
The data we have today shows that AI chatbots are developing into a significant customer acquisition channel with many familiar mechanics.
However, their task-based nature, bias, and demographics suggest we should be cautious when using the same approach as classic search engines.
Don’t forget – Search is just a means to an end. Ultimately, people search to solve problems, i.e., do tasks.
The fact that AI chatbots can skip the search part and do tasks on the spot means they’re superior to classic search engines. For this reason, I expect Google to add more agentic capabilities to AI Overviews or launch a new Gemini-based product in Search.
The underlying technology allows AI chatbots to fork off search engine ranks and develop their own signals. And it evolves rapidly.
The evolution so far went from machine learning in the pre-2022 era to early LLMs and now inference models (think: reasoning).
Better reasoning allows LLMs to recognize user intent even better than classic search engines, making it easier to train models on better sources to mention or cite.
This brings me to the question of whether Google/Bing incumbents will also dominate AI chatbots down the road. Right now, the answer is yes. But for how long?
Generational preferences could be the biggest driver of new platforms. The easiest way for Google to become irrelevant is to lose young people.
Semrush, 02/25: Searchers over 35 years use Google more often than ChatGPT. People between 18 and 24 use ChatGPT 46.7% of the time, compared to Google with 24.7%.
Higher Visibility, 02/25: 82% of Gen Z occasionally use AI chatbots, compared to 42% of Baby Boomers.
There is a chance that multimodality will quickly play a more prominent role in AI chatbot adoption. So far, text interfaces dominate.
But Google already reports 10 billion searches with Google Lens, and Meta’s Ray Ban smartglasses are very successful. Other than Google Search, the LLM answer format is easy to transport to other devices and modalities, which could transform AI.3
New research on AI Overviews and organic search results presents a fresh view on how AIO is evolving and suggests how to consider it for purposes of SEO.
Among the findings:
Their research showed that the AIO they were tracking showed more volatility than the organic search results, that they were changing at a faster rate.
AIO volatility doesn’t correlate with organic search volatility
They conclude that AIO is replacing featured snippets or “enhancing search results.”
It was also concluded that, for the purpose of SEO, AIO should be considered as something separate from the organic search.
Generative text changed for every query they looked at.
That last finding was really interesting and here is what they said about that:
“As far as I can tell, the generative text changed for every single query. However, our measure was looking for meaningful changes in the generative text which might reflect that Google had shifted the intent of the original query slightly to return different generative ranking pages.”
Another interesting insight was a caveat about search volatility is that it shouldn’t be taken as a sign of a Google update because it could be the influence of current events temporarily changing the meaning of a search query, which is related to Google’s freshness algorithm. I don’t know who the Authoritas people are but hats off to them, that’s a reasonable take on search volatility.
You can read the AIO research report here, it’s very long, so set aside at least 20 minutes to read it:
AIO Independence From Organic SERPs
That research published by Authoritas got me thinking about AIO, particularly the part about the AIO independence from the search results.
My thoughts on that finding is that there may be two reasons why AIO and organic SERPs are somewhat decoupled:
AIO is tuned to summarize answers to complex queries with data from multiple websites, stitching them together from disparate sources to create a precise long-form answer.
Organic search results offer answers that are topically relevant but not precise, not in the same way that AIO is precise.
Those are important distinctions. They explain why organic and AIO search results change independently. They are on independent parallel paths.
Those insights are helpful for making sense of how AIO fits into overall marketing and SEO strategies. Wrap your head around the insight that AIO and Organic Search do different and complementary things and AIO will seem less scary and become easier to focus on.
A complex query is something AIO can do better than the regular organic search results. An example of a complex question is asking “how” a general concept like men’s fashion is influenced by an unrelated factor like military clothing. Organic search falls short because Google’s organic ranking algorithm generally identifies a topically relevant answer and this kind of question demands a precise answer which may not necessarily exist on a single website.
What Is A Complex Query?
If complex queries trigger AI Overviews, where is the line? It’s hard to say because the line is moving. Google’s AIO are constantly changing. A short TL/DR answer could arguably be that adding a word like what or how can make a query trigger an AIO.
Example Of A Complex Query
Here’s the query:
“How is men’s fashion influenced by military style?”
Here’s the AIO answer that’s a summary based on information combined from from multiple websites:
“Men’s fashion is significantly influenced by military style through the adoption of practical and functional design elements like sharp lines, structured silhouettes, specific garments like trench coats, bomber jackets, cargo pants, and camouflage patterns, which originated in military uniforms and were later incorporated into civilian clothing, often with a more stylish aesthetic; this trend is largely attributed to returning veterans wearing their military attire in civilian life after wars, contributing to a more casual clothing culture.”
Here are the completely different websites and topics that AIO pulled that answer from:
Screenshot Of AIO Citations
The organic search results contain search results that are relevant to the topic (topically relevant), but don’t necessarily answer the question.
Information Gain Example
An interesting feature of AI Overviews is delivered through a feature that’s explained in a Google Patent on Information Gain. The patent is explicitly in the context of AI Assistants and AI Search. It’s about anticipating the need for additional information beyond the answer to a question. So in the example of “how is men’s fashion influenced by military style” there is a feature to show more information.
Screenshot Showing Information Gain Feature
The information gain section contains follow-up topics about:
Clean lines and structured fit
Functional design
Iconic examples of military clothing
Camouflage patterns
Post-war impact (how wars influenced what men after they returned home)
How To SEO For AIO?
I think it’s somewhat pointless to try to rank for information gain because what’s a main keyword and what’s a follow up question? They’re going to switch back and forth. Like, someone may query Google about the influence of camouflage patterns and one of the information gain follow-up questions may be about the influence of military clothing on camouflage.
The better way to think about AIO, which was suggested by the Authoritas study, is to just think about AIO as a search feature (which is what they literally are) and optimize for that in the same way one optimized for featured snippets, which in a nutshell is to create content that is concise and precise.
Hostinger announced a new service called Hostinger Horizons that allows anyone to build interactive online apps (like an AI-based website builder) without having to code or hire programmers. The new service allows users to turn their ideas into web applications by prompting an AI to create it.
AI Democratizes Entrepreneurship
In the early days of the Internet it seemed like people with backgrounds from Stanford University and Harvard Business School had access to the resources and connections necessary to turn ideas into functioning web apps. Over time, platforms like WordPress lowered the barrier to entry for starting and running online businesses, enabling virtually anyone to compete toe to toe with bigger brands. But there was still one last barrier and that was the ability to create web apps, the functionalities that power the biggest ideas on the Internet. Hostinger Horizons lowers that barrier, enabling anyone to turn their idea into a working app and putting entrepreneurial success within reach of anyone with a good idea. The significance of this cannot be overstated.
AI Powered Web App Builder
Hostinger Horizons is an AI-powered no-code platform created specifically for individuals and small businesses that enables them to create and publish interactive web applications without having to use third-party integrations or requiring programming knowledge.
The new platform works through an AI chat interface that creates what users are asking for while also showing a preview of the web app. A user basically prompts what they want, makes feature requests, tells it what to change and preview the results in real-time.
Hostinger Horizons speeds up the time it takes to create and deploy a functioning web app. Hosting and all other necessary services are integrated into the service, which simplifies creating web apps because there’s no need for third party services and APIs. Once an app is created an online a user can still return to it, edit and improve it in minutes. It promises to be a solution for fast prototyping without the technical and investment barriers that are typically associated with translating a good idea to deployment on the web.
The Hostinger announcement noted that simple web apps only takes minutes to create:
“Early access trials show that simple web apps, such as a personal calorie tracker, a language-learning card game, or a time management tool, can be built and published in minutes.”
How Hostinger Horizons Works
The new service combines AI-powered chat, with real-time previews and the ability to instantly publish the app to the web.
Hostinger provides all the necessary elements to get the work done:
Domain name registration
Email services
Multilingual support (80+ languages)
Supports image uploads
Supports user-provided sketches and screenshots
Voice prompting
Web hosting
Giedrius Zakaitis, Hostinger Chief Product and Technology Officer, offered these insights:
“Web apps have turned ideas into million-dollar startups, but building one always required coding or hiring a developer. We believe it is time to change the game. Just like Hostinger AI Website Builder introduced a new kind of site-building experience, Hostinger Horizons will democratize web apps so that anyone can bring their unique and exciting ideas online…”
Hostinger Horizons is an AI-powered no-code platform that is specifically designed to enable individuals and small businesses to build and publish fully functional web apps with no coding experience or external integrations needed. Users can just prompt what they want through an AI chat interface with real-time previews. It even allows uploading screenshots and sketches.
Hostinger Horizons promises to dramatically simplify the process of turning an idea into a working business by bundling hosting, domain registration, and email services into one solution.
Four reasons that make this a breakthrough service:
Rapid Prototyping: Create, modify, and deploy interactive apps in real-time, including rapid revisions after the app is published.
Integrated Services: Hosting and other essential tools are built in, eliminating reliance on third-party providers.
Democratized Development: Hostinger Horizons enables anyone to turn their ideas into an online business without technical barriers.
Supports 80+ languages
Creating Complex Websites With AI
What can you do with Hostinger Horizons? It seems like the right question to ask is what can’t you do with it. I asked Hostinger if the following applications of the technology was possible and they affirmed that the short answer is yes but that some of the ideas that I suggested may not be 100% straightforward to implement but that they were indeed possible to create.
Money makes the web run and I think applications that many would be interested in are ways to interactively engage users by enabling them to accomplish goals, capture leads, product comparison, improved shopping experiences and follow-up emails.
Since Hostinger Horizons handles hosting, domain registration, and email in a single platform, entrepreneurs and businesses can build these kinds of web pages by describing it to the AI chat interface, iteratively improving it and then publishing the finished project when it’s ready.
This could be useful to a restaurant, a law office, or a product review site, for example. Here are examples of the kinds of things I’d like to see it do.
Restaurant:
Reservation & Loyalty App Allows users to sign up and reserve tables and receive follow up reminders and offers.
Interactive Menu Explorer Can enable users to browse a menu according to dietary preferences and capture contact information for special offers.
Legal Office
Could be used to generate questionnaires and streamline the intake.
Product Reviews
Can encourage users to provide their requirements and preferences and then generate a summary of product reviews with quick links to where to purchase them.
Interactive Comparison Tools with links to where to purchase
As the top free app by downloads in the U.S. Apple app store since Jan. 26 – with 16 million app downloads in its first 18 days (ChatGPT had 9 million in the same timeframe) – DeepSeek’s performance and accompanying search feature is at least on par with OpenAI’s ChatGPT for a fraction of the cost.
U.S. tech analysts and investors seem to all fear that the U.S. is falling behind in the generative AI global race.
This may be warranted considering how quickly and cost-effectively DeepSeek was able to get R1 developed and out the door.
DeepSeek utilizes reinforcement learning, meaning the model learns complex reasoning behaviors through reinforcement without supervised fine-tuning, which allows it to save significant computational resources.
But, is DeepSeek really going to emerge as the leader in AI? And what are the implications for this development for the future of search? Let’s dive in.
What Has Happened Since DeepSeek Launched?
While U.S. tech companies were humbled by the speed and claimed cost efficiency of this launch, DeepSeek’s arrival has not been without controversy.
A lot of questions lurk, ranging from suspected intellectual property violations to security, data privacy, Chinese censorship, and the true cost of its technology.
Legal Issues For Copyright And Data Protection
OpenAI and Microsoft are investigating whether DeepSeek used OpenAI’s API to integrate their AI models into DeepSeek’s own models.
Distillation allows for the transfer of knowledge of a large pre-trained model into a smaller model, which enables the smaller model to achieve comparable performance to the large one while reducing costs.
This is more than a little ironic given the lawsuits against OpenAI for ignoring other site’s terms of service and using their copyrighted internet data to train its systems.
For anyone handling customer information and payment details, integrating a tool like DeepSeek that stores data in a foreign jurisdiction could violate data protection laws and expose sensitive information to unauthorized access.
Given that DeepSeek has yet to provide its privacy policies, industry experts and security researchers advise using extreme caution with sensitive information in DeepSeek.
It found a publicly accessible database belonging to DeepSeek, which allowed it full control over database operations and access to user data and API keys.
Wiz alerted the DeepSeek team, and they took immediate action to secure the data. However, it is unclear who else accessed or downloaded the data before it was secured.
While it’s not uncommon for startups to move fast and make mistakes, this is a particularly large mistake and shows DeepSeek’s lack of focus on cybersecurity so far.
National Security Concerns Similar To TikTok
There are national security concerns about DeepSeek’s data collection policies reminiscent of fears about TikTok, which saw a similar rise in global prominence out of Chinese-based company ByteDance.
The U.S. government briefly banned TikTok in January 2025, which came out of concerns about how the company was collecting data about users. There were also fears that the Chinese government could use the platform to influence the public in the U.S.
A few incidents in the last several years that initiated that fear include TikTok employees utilizing location data from the app to track reporters to find a source of leaked information, and TikTok employees being reported to have plans to track specific U.S. citizens.
While TikTok is active in the U.S. right now, its future is unconfirmed.
For similar reasons to the TikTok concerns, a number of governments around the world, including Australia and Italy, are already working to ban DeepSeek from government systems and devices. The U.S. is also considering a ban on DeepSeek.
Chinese Censorship
Regardless of whether you run DeepSeek locally or in its app, DeepSeek’s censorship is present for queries deemed sensitive by the Chinese government, according to a Wired investigation.
However, because it is open source, there are ways of getting around the censorship, but it’s difficult.
Doing so would require running on your own servers using modified versions of the publicly available DeepSeek code, which means you’d need access to several highly advanced GPUs to run the most powerful version of R1.
Questions About Cost
Much has been written about the cost of building DeepSeek. Initial claims by DeepSeek were that it took under $6 million to build based on the rental price of Nvidia’s GPUs.
However, a report from SemiAnalysis, a semiconductor research and consulting firm, has since argued that DeepSeek’s hardware spend was higher than $500 million, along with additional R&D costs.
For context, OpenAI lost about $5 billion in 2024 and anticipates it will lose more than $11 billion in 2025. Even if DeepSeek did cost $500 million or more, it still cut costs compared to what leading competitors are spending.
So, how did they cut costs?
Before DeepSeek came along, the leading AI technologies were built on neural networks, which are mathematical systems that learn skills by analyzing huge amounts of data. This requires large amounts of computing power.
Specialized computer chips called graphics processing units (GPUs) are an effective way to do this kind of data analysis. This is how chipmaker Nvidia grew to prominence (and also had a huge fall in market value on the day DeepSeek launched).
GPUs cost around $40,000 and require considerable electricity, which is why leading AI technologies like OpenAI’s ChatGPT were so expensive to build.
Sending data between chips can also require more energy than running the chips themselves.
Instead of creating one neural network that learned data patterns on the internet, they split the system into many neural networks and launched smaller “expert” systems paired with a “generalist” system, reducing the amount of data needed to travel between GPU chips.
The Implications Of Being Open Source
DeepSeek-R1 is as “open-source” as any LLM has been thus far, which means anyone can download, use, or modify its code.
Similar to Meta’s Llama, the code and technical explanations are shared, enabling developers and organizations to utilize the model for their own business needs, but the training data is not fully disclosed.
Many believe DeepSeek is a big step toward democratizing AI, allowing smaller companies and developers to build on DeepSeek-R1 and achieve greater AI feats faster.
This could lead to more innovation in places with more limited access to the tech needed to build AI solutions.
But, critics fear that open-source models can expose security vulnerabilities that could be exploited, which we’ve already seen in DeepSeek’s first weeks in the public.
DeepSeek And The Future of SEO
So, what does this all really mean for search professionals? The way I see it, DeepSeek is just the next splashy AI chatbot with search capabilities in the rapidly changing world of SEO.
It’s important to understand that while tools like DeepSeek and ChatGPT use advanced natural language processing (NLP) and machine learning, they still simply provide answers to real questions that real people ask.
Their responses heavily focus on semantic understanding, intent matching, and contextual analysis, but they ultimately serve the same core user need.
While we have years of experience testing optimization tactics on more established search engines like Google, we’re still at the beginning stages of understanding optimization for generative AI chatbots.
Final Thoughts
Whether DeepSeek will stick and grow in prominence remains to be seen.
Obviously, if other governments follow Australia, Italy, and potentially the U.S. to ban DeepSeek, that would limit its potential for growth.
And much as DeepSeek rose to prominence rapidly by providing a blueprint for others and significantly lowering costs, a new market-moving AI could always be just around the corner.
Regardless of what happens with DeepSeek, we are at the beginning of a very rapid period of innovation in AI technology.
As SEO professionals, we need to be prepared to test a surge of new platforms and reverse engineer how they arrive at their responses to user queries.
