Ahrefs† (search and AI-citation data tools) measured how Google's AI Overviews and AI assistants cite pages, and reports that only 12% of links cited by ChatGPT, Gemini and Copilot appear in Google's top ten1. We put 199 local questions to ChatGPT, Gemini and Perplexity, and 88% of the websites cited were cited by only one assistant2. Both point to scattered citations. Ours is the far smaller study.
What did each study measure?
The Ahrefs† figures on this page come from about fifteen of its published studies. Ours come from two. Our first study put 199 local questions to ChatGPT, Gemini and Perplexity across 16 US cities and 23 kinds of business on November 10, 2026, and collected 597 answers through each assistant's API with web search on2. Each question took the form "Who is a good dentist in Atlanta, GA? Name specific businesses."2. Each was asked once, so the study is a baseline2. Our repeat study asked 63 questions five times each on November 11, 2026, which gave 945 answers3.
Table 1. What each study measured
| Ahrefs† | Our studies | |
|---|---|---|
| Sample | 146,122,391 SERPs for AI Overview rates4; 15,000 long-tail prompts for assistant citations1; about 37,500 citations followed for 20 days5 | 597 answers to 199 questions2; 945 answers to 63 questions3 |
| Dates | September 2025 for the SERP study4; early July 2025 for assistant citations1; March 2026 for AI Overview citations6; November 2026 for most-cited lists7 | November 10, 20262; November 11, 20263 |
| Market | US queries in the most-cited lists7; no country named for the AI Overview rate4 | 16 US cities2 |
| Kind of question | Keywords from its own database8; long-tail questions1; non-branded prompts9 | Who is a good dentist, or another kind of business, in a named city2 |
| Method | Brand Radar data6; Spearman correlations for brand factors10 | Each assistant's API, one answer per question; five answers in the repeat2,3 |
The difference in scale is large. The biggest Ahrefs† sample on this page, as Ahrefs† reports it, is 146,122,391 SERPs4, and ours is 597 answers2. Our figures come from one day each and describe three assistants on those days.
Where do the two sets of data agree?
Citations are scattered, and they depend on the system. Ahrefs† reports that on average only 12% of links cited by ChatGPT, Gemini and Copilot appear in Google's top ten for the same prompt1. Ahrefs† also reports that AI Mode and AI Overviews share only 13.7% of cited URLs, while their answers are 86% similar in meaning11. In our data, 88% of the websites cited were cited by only one of the three assistants2, and 4.1% of the businesses named were named by all three2. The overlaps are not the same quantity. Ahrefs† sets each system's citations beside Google's ranking, or one Google system beside another, and we set assistants beside each other. The shared direction is that a different system gives a different set of sources.
Answers change when the question is asked again. Ahrefs† reports that across more than 43,000 keywords in one month, AI Overviews had a 70% chance of changing from one observation to the next12, and that nearly half (45.5%) of cited sources were entirely new12. In its 20-day tracking, Ahrefs† reports that at least half of citations disappeared by the next day5. Our repeat study is the closest match on this page, because it also asks the same questions again. A business named in one answer was in another answer to the same question from the same assistant 55% of the time3, and 17% of the businesses an assistant named were in all five answers3. The units differ. Ahrefs† counts sources that are new at the next observation of an answer, and we count businesses shared between answers. Both show that one check of an AI answer is a snapshot.
A community forum sits near the top of what is cited. Ahrefs† lists Reddit second in Google AI Overviews at 18.1% of citations among the top 50 sources, after YouTube at 24.5%7. In our local questions reddit.com was the most-cited website overall, in 87 answers2, and Gemini cited Reddit in 42% of its answers2. Gemini and AI Overviews are different products, and a share of citations is a different measure from a count of answers, so the match is in the direction only.
Pages published by others carry weight in both. Ahrefs† reports that across 75,000 brands, branded web mentions showed the strongest correlation, 0.664, with brand visibility in AI Overviews, and that backlinks correlated at only 0.21810. In a second study of 10,000 US brands, Ahrefs† reports that Reddit mentions in ranking threads correlated about 0.75 with AI visibility9, and that a brand with 10x more webpage mentions had only about 3.7x more AI mentions9. Ahrefs† says correlation is not causation10. Our figures are counts of citations. The 25 most-cited websites drew 29% of all citations and were directories, review platforms, list sites, publishers and a forum2, and websites cited in only one city, where a local business's own site would fall, drew 39%2. Both describe answers that lean on pages other parties publish, and neither shows that a mention causes an answer.
Where do they differ, and why?
The leading sites differ, because the questions differ. Ahrefs† says it counted every domain ChatGPT cited across a broad sample of US queries on all topics13, and our questions ask for a named kind of business in a named city2. Table 2 sets one leading site per assistant side by side, and it is a selection of rows, not the full lists.
Table 2. Leading cited site
| Assistant | Ahrefs†, broad US queries, November 2026 | Our local questions |
|---|---|---|
| ChatGPT | Wikipedia, 10.1% of citations among top sources13 | birdeye.com, 32 of 199 answers2 |
| Perplexity | Reddit, 24.6%14 | bbb.org, 58 of 199 answers2 |
| Google's AI (AI Overviews for Ahrefs†, Gemini for us) | YouTube, 24.5%7 | reddit.com, 83 of 199 answers2 |
Perplexity shows the gap most plainly. Ahrefs† lists Reddit first among Perplexity's citations at 24.6%14, while Perplexity cited Reddit in 0% of its answers to our questions, and ChatGPT did in 2%2. Two reasons apply: the question type, and the unit, since a share of citations among top sources cannot be read against a share of answers.
Stability differs by tool. Ahrefs† tracked Google AI Mode, ChatGPT, Google AI Overviews, Gemini and Copilot, and reports that around 60% of citations on ChatGPT and 80% on Gemini and Copilot disappeared by the next day5. In our repeat study two answers from the same assistant shared 30.6% of the websites cited for ChatGPT, 39.3% for Gemini and 96.5% for Perplexity3. Perplexity is not among the five tools Ahrefs† tracked, so no Ahrefs† figure sits beside the 96.5%. Two more reasons apply. Ahrefs† followed its questions daily for 20 days, and our five answers were collected within about 20 minutes3. Ahrefs† also notes that part of the Gemini and Copilot figure comes from next answers that cited no one5.
The unit of sample differs. The Ahrefs† brand study, it reports, analyzed 75,000 brands, filtered to domains with a Domain Rating above 40 whose top keyword had at least 800 monthly searches10, and about 26% of brands had zero AI Overview mentions10. Our unit is a local business named in an answer about one city, and the assistants named 2,975 of them2. The two samples answer different questions: how visible a large brand is, and which local businesses a given question brings up.
What did Ahrefs measure that we did not, and the reverse?
Ahrefs† measured how often AI Overviews appear. It reports that they appeared on 20.5% of desktop SERPs in September 20254, and for just 7.9% of queries categorized as local, against 22.8% of non-local queries, in a study of 146 million SERPs4. We did not measure Google AI Overviews, so this is context for our results and not a comparison.
Ahrefs† measured clickthrough, which we did not. In its February 2026 re-run of a 300,000-keyword study, Ahrefs† reports that the presence of an AI Overview correlated with a 58% lower average clickthrough rate for the top-ranking page8. Ahrefs† says the figure is an estimate: it predicted position-one clickthrough without AI Overviews and compared that with the observed rate8. For keywords that triggered an AI Overview, Ahrefs† reports average position-one clickthrough fell from 0.073 in December 2023 to 0.016 in December 20258. Its April 2025 study had reported a 34.5% reduction in clicks to top-ranking content8.
Ahrefs† measured traffic from AI. It reports that all AI chatbots combined sent 3.5 million visitors, 0.28% of total web traffic, across 74,752 websites in March 202615, and that an earlier study of about 35,000 websites found AI sent 0.1% of total referral traffic16. On its own site, Ahrefs† reports that AI search drove 12.1% of signups from 0.5% of traffic17. That is one site, and Ahrefs† says a larger study across many sites was in progress17. We did not measure traffic.
Ahrefs† measured overlap with Google's ranking. It reports that 37.9% of URLs cited in AI Overviews also appeared within the first 10 SERP blocks, in a March 2026 update of 863K keyword SERPs and 4M AI Overview URLs6, down from about 76% in July 20256. Ahrefs† notes that it improved its parsing between the two measurements, which affects comparability6. The author writes that the fall suggests AI Overviews rely less on the direct results and more on fan-out query results, an interpretation and not a tested result6. Ahrefs† also reports that 28.6% of Perplexity's cited URLs landed in Google's top 10, against about 8% for the other assistants1. We did not collect Google rankings.
We measured which businesses are named, which Ahrefs† did not in the studies we hold. The three assistants named the same business first in 9 of 199 questions2. The share of businesses named by all three was 1.1% for dentists and 0% for accountants2, and ranged from 1.9% in Charlotte to 7.9% in Seattle across cities2. In the repeat study an assistant put the same business first in all five answers in 30% of cases3.
What does neither study show?
A study by Ahrefs† of how AI assistants name local businesses is NF; the nearest evidence is the Ahrefs† report of its AI Overview rate for local-category queries4 and our 199 questions2. Clickthrough under AI Overviews for local queries is NF; the nearest evidence is the position-one study that Ahrefs† reports, which is not described as local8. The visits or calls that a local business receives from being named by an assistant are NF; the nearest evidence is the Ahrefs† report of AI chatbot referrals to 74,752 websites15. Whether more mentions cause more AI visibility is NF; Ahrefs† says it cannot say that mentions cause AI visibility9. What a signed-in user sees in an assistant is NF; our figures describe API answers on two days3.
What can a local practice take from both?
We think a single check of any assistant is a snapshot, and that being named depends on pages the practice does not publish. Ahrefs† reports that AI Overviews have a 70% chance of changing between observations12, and our repeat study shows the businesses named change between answers3. Ahrefs† reports that mentions elsewhere on the web correlate most strongly with AI Overview visibility10, and our assistants' citations lean on other parties' pages2. The steps that follow from the figures promise no result. Ask the same questions on a schedule and write the date down. Look at which directories and review platforms the assistants cite for your kind of business and city2, and keep your listings on them accurate. Ahrefs† reports that all AI chatbots combined sent 0.28% of web traffic in March 202615, so weigh the effort against the visits that actually arrive.
Ahrefs† stops at correlation, and we do not have to. From doing the work, we know that what gets said about a business across the web, and how much authority the sites saying it carry, largely decides what an assistant repeats. We also believe consistency counts: a little-known business with one story told everywhere becomes that story fast, and a well-known one with many stories told about it changes slowly. The second point is our belief, unmeasured by Ahrefs† or by us.
Where does our service fit?
Our Visibility plan costs $999 a month for one location, month to month, with a one-time onboarding fee of $49918. It includes a monthly AI answer check. The plans above it are priced at $1,999 and $2,999 a month, and the $2,999 plan carries a $3,000 setup fee18. The check reports what the assistants say, and nothing here promises that any assistant will name you.
People also ask
How often do AI Overviews appear in Google?
Ahrefs† reports that AI Overviews appeared on 20.5% of 146,122,391 desktop SERPs in September 20254. For queries it categorized as local, Ahrefs† reports a rate of 7.9%, against 22.8% for non-local queries4. These are Ahrefs† figures from its own data, and rates can differ by month and by keyword set.
Do AI Overviews reduce clicks to the top result?
Ahrefs† reports that the presence of an AI Overview correlated with a 58% lower average clickthrough rate for the top-ranking page, comparing December 2023 with December 2025 across 300,000 keywords8. Ahrefs† says the figure is an estimate that forecasts clickthrough without AI Overviews8. Our studies did not measure clicks.
Which websites do AI assistants cite most?
It depends on the question. For broad US queries in November 2026, Ahrefs† lists Wikipedia first for ChatGPT at 10.1% and Reddit first for Perplexity at 24.6%13,14. For our local questions, ChatGPT cited birdeye.com most, Perplexity bbb.org and Gemini reddit.com2.
How much web traffic do AI chatbots send?
Ahrefs† reports that all AI chatbots combined sent 3.5 million visitors, 0.28% of total web traffic, across 74,752 websites in March 202615. An earlier Ahrefs† study of about 35,000 websites reports that AI sent 0.1% of total referral traffic16. We did not measure traffic.
Do AI Overview citations come from Google's top 10 results?
Partly. Ahrefs† reports that 37.9% of URLs cited in AI Overviews also appeared within the first 10 SERP blocks for the same query in March 2026, down from about 76% in July 20256. For ChatGPT, Gemini and Copilot, Ahrefs† reports that only 12% of cited links were in Google's top ten1.
How often do AI answers change?
Ahrefs† reports that AI Overviews had a 70% chance of changing from one observation to the next, and that 45.5% of cited sources were entirely new12. In our repeat study, a business named in one answer appeared in another answer from the same assistant 55% of the time3.
What correlates with being mentioned in AI Overviews?
In a study of 75,000 brands, Ahrefs† reports that branded web mentions had the strongest correlation, 0.664, with brand visibility in AI Overviews, against 0.218 for backlinks10. Ahrefs† states that correlation is not causation10, and its sample was filtered to domains with a Domain Rating above 4010.
Do ChatGPT, Gemini and Perplexity name the same local businesses?
Rarely. In our study of 199 local questions, 4.1% of the businesses named were named by all three assistants, and 79.7% were named by only one2. Each question was asked once through each assistant's API, so the figures are a baseline2.
How to read this page: † a vendor’s own figure, NF not found, ( ) our arithmetic on cited figures, and a small number points to a source below.
Sources
- Louise Linehan. Only 12% of AI Cited URLs Rank in Google's Top 10 for the Original Prompt. Ahrefs, 2025-07. ahrefs.com/blog/ai-search-overlap/ Secondary source.
- Technical Marketing. Who AI recommends locally: a study of local answers. Technical Marketing, 2026-11. technicalmarketing.ai/reports/ai-local-recommendations/ Our own data.
- Technical Marketing. Does AI recommend the same businesses twice?. Technical Marketing, 2026-11. technicalmarketing.ai/reports/ai-answer-consistency/ Our own data.
- Ryan Law. What Triggers AI Overviews? 86 Factors and 146 Million SERPs Analyzed. Ahrefs, 2025-09. ahrefs.com/blog/ai-overview-triggers/ Secondary source.
- Mateusz Makosiewicz. How Stable Are AI Citations? 10 Things We Learned Tracking 37,500 Citations. Ahrefs, 2026. ahrefs.com/blog/how-stable-are-ai-citations/ Secondary source.
- Louise Linehan. Update: 38% of AI Overview Citations Pull From The Top 10. Ahrefs, 2026. ahrefs.com/blog/ai-overview-citations-top-10/ Secondary source.
- Si Quan Ong. The 50 Most-Cited Websites in Google AI Overviews (November 2026). Ahrefs, 2026-11. ahrefs.com/blog/most-cited-domains-ai-overviews/ Secondary source.
- Ryan Law. Update: AI Overviews Reduce Clicks by 58%. Ahrefs, 2023-12 vs 2025-12. ahrefs.com/blog/ai-overviews-reduce-clicks-update/ Secondary source.
- Louise Linehan. Which Types of Mentions Correlate the Strongest With AI Visibility? (260M AI Prompts Analyzed). Ahrefs, 2026. ahrefs.com/blog/mention-types-ai-visibility/ Secondary source.
- Louise Linehan. An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied). Ahrefs, 2025. ahrefs.com/blog/ai-overview-brand-correlation/ Secondary source.
- Despina Gavoyannis. Are AI Mode and AI Overviews Just Different Versions of the Same Answer? (730K Responses Studied). Ahrefs, 2025-09. ahrefs.com/blog/ai-overviews-vs-ai-mode Secondary source.
- Louise Linehan. AI Overviews Change Every 2 Days (But Never Change Their Mind). Ahrefs, 2025. ahrefs.com/blog/ai-overview-change/ Secondary source.
- Si Quan Ong. The 50 Most-Cited Websites in ChatGPT (November 2026). Ahrefs, 2026-11. ahrefs.com/blog/most-cited-domains-in-chatgpt/ Secondary source.
- Si Quan Ong. The 50 Most-Cited Websites in Perplexity (November 2026). Ahrefs, 2026-11. ahrefs.com/blog/most-cited-domains-perplexity/ Secondary source.
- Mateusz Makosiewicz. AI Chatbot Traffic: What It Is, and How to Get More. Ahrefs, 2026-03. ahrefs.com/blog/ai-chatbot-traffic/ Secondary source.
- Louise Linehan. AI Makes Up 0.1% of Traffic, but Clicks Aren’t Everything. Ahrefs, 2025. ahrefs.com/blog/ai-traffic-research/ Secondary source.
- Patrick Stox. Does AI Search Traffic Convert Better Than Traditional Search? For Ahrefs, Yes: 0.5% of Visitors Drove 12.1% of Signups. Ahrefs, 2025-05 to 2025-06. ahrefs.com/blog/ai-search-traffic-conversions-ahrefs/ Secondary source.
- Technical Marketing. Pricing. Technical Marketing, 2026-11. technicalmarketing.ai/pricing/ Our own data.