Google Links To Itself: 43% Of AI Overviews Point Back To Google via @sejournal, @MattGSouthern

New research shows that Google’s AI Overviews often link to Google, contributing to the walled garden effect that encourages users to stay longer on Google’s site.

A study by SE Ranking examined Google’s AI Overviews in five U.S. states. It found that 43% of these AI answers contain links redirecting users to Google’s search results. Each answer typically includes 4-6 links to Google.

This aligns with recent data indicating that Google users make 10 clicks before visiting other websites. These patterns suggest that Google is working to keep users within its ecosystem for longer periods.

Google Citing Itself in AI Answers

The SE Ranking study analyzed 100,013 keywords across five states: Colorado, Texas, California, New York, and Washington, D.C.

It tracked how Google’s AI summaries function in different regions. Although locations showed slight variance, the study found that Google.com is the most-cited website in AI Overviews.

Google appears in about 44% of all AI answers, significantly ahead of the next most-cited sources, YouTube, Reddit, Quora, and Wikipedia, appearing in about 13%.

The research states:

“Based on the data combined from all five states (141,507 total AI Overview appearances), our data analysis shows that 43.42% (61,437 times) of AI Overview responses contain links to Google organic results, while 56.58% of responses do not.”

Image Credit: SE Ranking

Building on the Walled Garden Trend

These findings complement a recent analysis from Momentic, which found that Google’s “pages per visit” has reached 10, indicating users spend significantly more clicks on Google before visiting other sites.

Overall, this research reveals Google is creating a more self-contained search experience:

• AI Overviews appear in approximately 30% of all searches
• Nearly half of these AI answers link back to Google itself
• Users now make 10 clicks within Google before leaving
• Longer, more specific queries trigger AI Overviews more frequently

Google still drives substantial traffic outward; 175.5 million visits in March, according to Momentic.

However, it’s less effective at sending users away than ChatGPT. Google produces just 0.6 external visits per user, while ChatGPT generates 1.4 visits per user.

More Key Stats from the Study

The SE Ranking research uncovered several additional findings:

• AI Overviews almost always appear alongside other SERP features (99.25% of the time), most commonly with People Also Ask boxes (98.5%)
• The typical AI Overview consists of about 1,766 characters (roughly 254 words) and cites an average of 13.3 sources
• Medium-difficult keywords (21-40 on the difficulty scale) most frequently trigger AI Overviews (33.4%), whereas highly competitive terms (81-100) rarely generate them (just 3.7%)
• Keywords with CPC values between $2-$5 produce the highest rate of AI Overviews (32%), while expensive keywords ($10+) yield them the least (17.3%)
• Fashion and Beauty has the lowest AI Overview appearance rate (just 1.4%), followed by E-Commerce (2.1%) and News/Politics (3.8%)
• The longer an AI Overview’s answer, the more sources it cites. Responses under 600 characters cite about five sources, while those over 6,600 characters cite around 28 sources.

These statistics further emphasize how Google’s AI Overviews are reshaping search behavior.

This data stresses the need to optimize for multiple traffic sources while remaining visible within Google’s results pages.

U.S. Copyright Office Cites Legal Risk At Every Stage Of Generative AI via @sejournal, @martinibuster

The United States Copyright Office released a pre-publication version of a report on the use of copyrighted materials for training generative AI, outlining a legal and factual case that identifies copyright risks at every stage of generative AI development.

The report was created in response to public and congressional concern about the use of copyrighted content, including pirated versions, by AI systems without first obtaining permission. While the Copyright Office doesn’t make legal rulings, the reports it creates offer legal and technical guidance that can influence legislation and court decisions.

The report offers four reasons AI technology companies should be concerned:

1. The report states that many acts of data acquisition, the process of creating datasets from copyrighted work, and training could “constitute prima facie infringement.”
2. It challenges the common industry defense that training models does not involve “copying,” noting that the process of creating datasets involves the creation of multiple copies, and that improvements in model weights can also contain copies of those works. The report cites reports of instances where AI reproduces copyrighted works, either word for word or “near identical” copies.
3. It states that the training process implicates the right of reproduction, one of the exclusive rights granted to emphasizes that memorization and regurgitation of copyrighted content by models may constitute infringement, even if unintended.
4. Transformative use, where it adds a new meaning to an original work, is an important consideration in fair use analysis. The report acknowledges that “some uses of copyrighted works in AI training are likely to be transformative,” but it “disagrees” with the argument that AI training is transformative simply because it resembles “human learning,” such as when a person reads a book and learns from it.

Copyright Implications At Every Stage of AI Development

Perhaps the most damning part of the report is where it says that there may be copyright issues at every stage of the AI development and lists each stage of development and what may be wrong with it.

“A. Data Collection and Curation

The steps required to produce a training dataset containing copyrighted works clearly implicate the right of reproduction…

B. Training

The training process also implicates the right of reproduction. First, the speed and scale of training requires developers to download the dataset and copy it to high-performance storage prior to training.96 Second, during training, works or substantial portions of works are temporarily reproduced as they are “shown” to the model in batches.

Those copies may persist long enough to infringe the right of reproduction,160 depending on the model at issue and the specific hardware and software implementations used by developers.

Third, the training process—providing training examples, measuring the model’s performance against expected outputs, and iteratively updating weights to improve performance—may result in model weights that contain copies of works in the training data. If so, then subsequent copying of the model weights, even by parties not involved in the training process, could also constitute prima facie infringement.

C. RAG

RAG also involves the reproduction of copyrighted works.110 Typically, RAG works in one of two ways. In one, the AI developer copies material into a retrieval database, and the generative AI system can later access that database to retrieve relevant material and supply it to the model along with the user’s prompt.111 In the other, the system retrieves material from an external source (for example, a search engine or a specific website).181 Both methods involve making reproductions, including when the system copies retrieved content at generation time to augment its response.

D. Outputs

Generative AI models sometimes output material that replicates or closely resembles copyrighted works. Users have demonstrated that generative AI can produce near exact replicas of still images from movies,112 copyrightable characters,113 or text from news stories.114 Such outputs likely infringe the reproduction right and, to the extent they adapt the originals, the right to prepare derivative works.”

The report finds infringement risks at every stage of generative AI development, and while its findings are not legally binding, they could be used to create legislation and serve as guidance for courts.

Takeaways

• AI Training And Copyright Infringement:
  The report argues that both data acquisition and model training can involve unauthorized copying, possibly constituting “prima facie infringement.”
• Rejection Of Industry Defenses:
  The Copyright Office disputes common AI industry claims that training does not involve copying and that AI training is analogous to human learning.
• Fair Use And Transformative Use:
  The report disagrees with the broad application of transformative use as a defense, especially when based on comparisons to human cognition.
• Concern About All Stages Of AI Development:
  Copyright concerns are identified at every stage of AI development, from data collection, training, retrieval-augmented generation (RAG), and model outputs.
• Memorization and Model Weights:
  The Office warns that AI models may retain copyrighted content in weights, meaning even use or distribution of those weights could be infringing.
• Output Reproduction and Derivative Works:
  The ability of AI to generate near-identical outputs (e.g., movie stills, characters, or articles) raises concerns about violations of both reproduction and derivative work rights.
• RAG-Specific Infringement Risk:
  Both methods of RAG, copying content into a database or retrieving from external sources, are described as involving potentially infringing reproductions.

The U.S. Copyright Office report describes multiple ways that generative AI development may infringe copyright law, challenging the legality of using copyrighted data without permission at every technical stage, from dataset creation to model outputs. It rejects the use of the analogy of human learning as a defense and the industry’s broad application of fair use. Although the report doesn’t have the same force as a judicial finding, the report can be used as guidance for lawmakers and courts.

Featured Image by Shutterstock/Treecha

New AI Models Make More Mistakes, Creating Risk for Marketers via @sejournal, @MattGSouthern

The newest AI tools, built to be smarter, make more factual errors than older versions.

As The New York Times highlights, tests show errors as high as 79% in advanced systems from companies like OpenAI.

This can create problems for marketers who rely on these tools for content and customer service.

Rising Error Rates in Advanced AI Systems

Recent tests reveal a trend: newer AI systems are less accurate than their predecessors.

OpenAI’s latest system, o3, got facts wrong 33% of the time when answering questions about people. That’s twice the error rate of their previous system.

Its o4-mini model performed even worse, with a 48% error rate on the same test.

For general questions, the results (PDF link) were:

• OpenAI’s o3 made mistakes 51% of the time
• The o4-mini model was wrong 79% of the time

Similar problems appear in systems from Google and DeepSeek.

Amr Awadallah, CEO of Vectara and former Google executive, tells The New York Times:

“Despite our best efforts, they will always hallucinate. That will never go away.”

Real-World Consequences For Businesses

These aren’t just abstract problems. Real businesses are facing backlash when AI gives wrong information.

Last month, Cursor (a tool for programmers) faced angry customers when its AI support bot falsely claimed users couldn’t use the software on multiple computers.

This wasn’t true. The mistake led to canceled accounts and public complaints.

Cursor’s CEO, Michael Truell, had to step in:

“We have no such policy. You’re of course free to use Cursor on multiple machines.”

Why Reliability Is Declining

Why are newer AI systems less accurate? According to a New York Times report, the answer lies in how they’re built.

Companies like OpenAI have used most of the available internet text for training. Now they’re using “reinforcement learning,” which involves teaching AI through trial and error. This approach helps with math and coding, but seems to hurt factual accuracy.

Researcher Laura Perez-Beltrachini explained:

“The way these systems are trained, they will start focusing on one task—and start forgetting about others.”

Another issue is that newer AI models “think” step-by-step before answering. Each step creates another chance for mistakes.

These findings are concerning for marketers using AI for content, customer service, and data analysis.

AI content with factual errors could hurt your search rankings and brand.

Pratik Verma, CEO of Okahu, tells the New York Times:

“You spend a lot of time trying to figure out which responses are factual and which aren’t. Not dealing with these errors properly basically eliminates the value of AI systems.”

Protecting Your Marketing Operations

Here’s how to safeguard your marketing:

• Have humans review all customer-facing AI content
• Create fact-checking processes for AI-generated material
• Use AI for structure and ideas rather than facts
• Consider AI tools that cite sources (called retrieval-augmented generation)
• Create clear steps to follow when you spot questionable AI information

The Road Ahead

Researchers are working on these accuracy problems. OpenAI says it’s “actively working to reduce the higher rates of hallucination” in its newer models.

Marketing teams need their own safeguards while still using AI’s benefits. Companies with strong verification processes will better balance AI’s efficiency with the need for accuracy.

Finding this balance between speed and correctness will remain one of digital marketing’s biggest challenges as AI continues to evolve.

Featured Image: The KonG/Shutterstock