Data Studies
Research on how ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews answer the questions your customers ask: which brands they name, which pages they cite and link to, and where they send shoppers. Most studies are built on thousands of answers we collect from those engines, including the web searches ChatGPT runs behind the scenes. Many compare those answers with Google search and Google Maps results, Google Business Profiles, Google News, or robots.txt files from the public HTTP Archive crawl. Use them to see where AI answers pull from in your market.
122 studies · page 1 of 6 · Need help acting on them? See our AI SEO services.
The first website ChatGPT cites in an answer is the same website Google ranks #1 for that question 20.2% of the time, so four times out of five ChatGPT leads with a different source. Agreement runs from 69.4% on software questions down to 0.5% on recipe questions.
87.7% of the pages ChatGPT cites in its answers are not in Google's top 10 results for the same question. Even counting whole websites, 66.2% of the websites ChatGPT cites have no page on Google's first page, and in 41.9% of ChatGPT answers not one cited website appears there.
57.6% of Google searches for products and services now open with an AI Overview, the AI-written answer Google places above the regular results. On phones the rate is higher: 61.2% against 55.4% on a computer for the same keywords.
On 40.2% of Google searches made on a phone, the first regular website listing starts below the bottom of the screen, so the searcher has to scroll before seeing a single website. Map listings, not AI Overviews, do most of the pushing.
Ask ChatGPT, Gemini and Perplexity the same "what is the best..." question and all three name the same brand first on 25.5% of questions. Three times in four, at least one AI assistant puts a different brand at the top, and no two assistants agree even half the time.
Google's own AI assistant, Gemini, leads its answer with the website Google Search ranks #1 for the same question in 9.5% of answers, so nine times in ten Google's two products point to different sources. ChatGPT matches Google's #1 result twice as often, at 20.2%.
82.7% of the pages Perplexity cites in its answers are not in Google's top 10 results for the same question. Perplexity lists 18.4 sources per answer on average, nearly twice the ten regular results on Google's first page, and most of them come from beyond it.
14.4% of the local businesses ChatGPT or Gemini describe as open 24/7 do not list round-the-clock hours on their own Google Business Profile. Phone numbers hold up better but not perfectly: 1.4% of ChatGPT's local business listings show a number that does not match Google, rising to 3.9% for emergency plumbers.
One in 383 local business recommendations from ChatGPT and Gemini that match a Google Business Profile is a business Google marks as closed. A bigger gap is matching: 1,350 of 10,000 recommended businesses have no Google profile at all, and 31% of the business names Gemini gives in its answer text match a Google Business Profile, against 94.8% of ChatGPT's picks.
1.8% of the links ChatGPT cites lead to a page that no longer exists or a server error, against 0.5% for Gemini and Google AI Mode. Across all three AI engines, 0.9% of cited links are dead, and one cited page in six turns away automated visitors.
38% of AI answers to questions about the latest, newest or current version of something cite at least one page more than a year old, and 24.3% cite a page more than two years old. The reader asked for what is new; more than a third of answers lean partly on last year's information.
Ask ChatGPT the same "what is the best..." question twice in a row and it names a different top pick 56.3% of the time; Gemini changes its top pick 82.7% of the time. Saying "please", swapping "best" for "top", making a typo or asking in Spanish adds only a few points to that everyday churn, while adding a budget changes ChatGPT's top pick 87.4% of the time.
Only 3.9% of the local businesses ChatGPT recommends are chains, against 6.3% of the top 20 Google Maps results for the same searches. The gap is widest for locksmiths: chains hold 29% of Google Maps' top 20 but 2.6% of ChatGPT's locksmith picks.
ChatGPT recommends 1.9% of local businesses with fewer than 10 Google reviews, against 26.9% of businesses with 500 to 999 reviews, a 14-fold gap. Nearly a third of businesses on Google Maps sit below that 10-review line.
88.4% of the local businesses ChatGPT recommends are not among the top three Google Maps results for the same service and city, the three businesses Google shows in its map pack. 17.5% are not in Google Maps' top 100 at all.
17.4% of the local businesses ChatGPT and Gemini recommend for a US city are located in a different city from the one in the question. In Cleveland and Detroit, more than half of AI's local picks are outside the city.
Yelp appears among ChatGPT's sources in 0.2% of its answers about the best local businesses, 3 answers out of 1,580, while Expertise.com appears in 28% and Angi in 20.6%. Gemini skips directories almost entirely and links Google Maps in 71.5% of its local answers.
Ask ChatGPT for the best local business in a US city and it names the highest-rated one on Google Maps only 46.2% of the time, so more than half the time the best-reviewed business is left out. Gemini names it 8.5% of the time.
In Cleveland, only 37.3% of the local businesses ChatGPT recommends are real, open and inside the city, against 93.8% in San Antonio. Nationwide, 72.5% of ChatGPT's local recommendations check out, so more than one in four fail at least one of the checks.
ChatGPT recommends 12.3% of local businesses that have a website on their Google Maps listing and 3.2% of those without one, a 3.9x gap. On Gemini the gap widens to 10.5x.
ChatGPT links to Amazon in 0.1% of its answers to "what is the best..." shopping questions, four answers in 4,000, while Perplexity links to Amazon in 17.4%. Across five AI engines, 5.5% of shopping answers link to Amazon.
For the same "best product" search, the products ChatGPT recommends cost 133% more than Google Shopping's top 10 results in the typical case, and they cost more in 38 of 41 searches.
Walmart receives 16.2% of links to US retailers in AI shopping answers and Target 9.9%, against 11.8% for Amazon. ChatGPT sends 23.9% of its retailer links to Walmart and 12.6% to Target.
ChatGPT shows a star rating on 2.8% of the product cards it puts in shopping answers, so almost every product it recommends reaches shoppers with no score attached. When a rating does appear, it is usually a strong one: 4.6% of rated recommendations sit under 4 stars.
Make Search Your Best Channel
Get a tailored growth plan for your site, backed by a senior team with 20+ years shipping results.
Get Your Free Proposal