How AI reads this page
ChatGPT, Claude, Gemini and Perplexity don't look at colours, images or animations. They receive text, headings, links and structured data. Below is exactly what they read when they open “Where does ChatGPT traffic land? 28.8% ends up in your site search”.
The AI receives 1,976 words in 9 sections, 13 links and 3 blocks of structured data.
Text
The page content in Markdown, without menus or graphics. It is the format closest to how a language model reads text.
# Where does ChatGPT traffic land? 28.8% ends up in your site search > Previsible's study of 6.77 million sessions: ChatGPT sends 28.8% of visits to internal search, then to product and About pages. What to check and how to measure it. URL: https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare Language: English Version IT: https://geosnap.ai/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare Author: Rinald Sefa, CMO Geosnap · Category: Guides Publisher: Geosnap (Maind Group S.r.l.), https://geosnap.ai ## In short Visits from ChatGPT do not arrive on blog posts: according to Previsible (6.77 million sessions, 166 sites, November 2024 to May 2026) 28.8% land on the internal search results page, then on product pages (43% in e-commerce), course pages (52%) and About pages (42.1% in health). It happens because the assistant trusts the domain but does not know which page to pick. The fix is to test your site search with the ten most frequent questions, remove empty results pages, push decision pages to the top and measure AI traffic by page type in GA4. - 28.8% of the visits ChatGPT sends land on the site's internal search (34.6% in SaaS); roughly a quarter across all AI assistants combined. - The other entry points are the pages that close a decision: product pages (43% in e-commerce), courses (52%), About pages (42.1% in health). - ChatGPT builds a link to your search with the user's question because it trusts the domain but not the page: structural behaviour, not a glitch. - Test your internal search with the ten most frequent questions every quarter: no empty pages, results sorted by intent, readable titles, parameter kept in the URL. - In GA4 measure AI traffic by page type (AI sessions divided by total sessions) and switch on site search in enhanced measurement; remember ChatGPT sometimes drops UTMs. Written by the Geosnap AI agent, reviewed and approved by Rinald Sefa. Translated from the Italian original. [Read the original](https://geosnap.ai/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare) When ChatGPT recommends your site and someone clicks, where do they land? Almost never on a blog post. Previsible's study of 6.77 million sessions referred by AI assistants shows that 28.8% of the visits ChatGPT sends arrive on the site's internal search results page, and that the rest cluster on product pages, course pages, About pages and contact pages. These are the pages nobody looks after as entry points. Here are the numbers, the reason behind the behaviour, and a checklist you can run in an afternoon. ## Which pages do ChatGPT visitors land on? Functional pages, not editorial ones. The [Previsible 2026 report](https://previsible.com/seo-strategy/ai-traffic-report-july-2026/), published on 6 July 2026, analyses 6.77 million sessions from ChatGPT, Perplexity, Gemini, Copilot and Claude across 166 Google Analytics 4 properties, from November 2024 to May 2026, in nine sectors (SaaS, e-commerce, finance, legal, health, insurance, education, publishing, ticketing). ChatGPT accounts for 92.4% of those sessions, so its habits shape everything else. The breakdown by page type looks like this: | Sector | Most frequent landing page | Share of AI traffic | | --- | --- | --- | | All (ChatGPT only) | Internal search results | 28.8% | | SaaS | Internal search results | 34.6% | | E-commerce | Product pages | 43% | | Education | Course pages | 52% | | Health | About page | 42.1% | | Publishing | News pages | 54% | | Legal | Blog 28.2%, About 12.3%, contact 11.7%, locations 10.3% | — | | Ticketing | Internal search 20.8%, home 18.5% | — | Overall, roughly a quarter of the visits sent by AI assistants end up on an internal search page. Perplexity behaves differently: it sends 13% of classified traffic to blog pages, against an 8.7% average. Previsible separates "search-pattern" models (ChatGPT and Gemini: they trust the domain but are unsure which page to pick) from "content-selection" models (Perplexity and Claude: they choose the exact article). Other measurements give different numbers because they use different methods. Similarweb, working from a browsing panel rather than the sites' own Analytics data, estimates that 58.8% of ChatGPT clicks land on the homepage, a share that roughly doubled once ChatGPT started showing more prominent brand links, as [Search Engine Journal](https://www.searchenginejournal.com/ai-referrals-are-recreating-the-oldest-mistake-in-conversion-optimization/584331/) reports. The two studies agree on one thing: people arrive on service pages (home, search, product), not on the content that got the site cited. ## Why does ChatGPT send people to your internal search? Because it trusts your domain but is not sure about the page. When the assistant has worked out that the answer lives on your site but has no precise, current URL, it builds a link to your internal search with the user's question as a parameter, in the form `yoursite.com/?s=keyword`. Previsible describes this as behaviour that "appears structural", not a passing glitch: in SaaS it accounts for a third of visits. The result is a visitor who arrives already informed (they have read an answer, they know what they want) and meets the page the site cares about least: no heading, results sorted by date, no filters, often a "no results" message because the internal engine cannot handle natural-language questions. [Slobodan Manic](https://www.searchenginejournal.com/ai-referrals-are-recreating-the-oldest-mistake-in-conversion-optimization/584331/) calls it a repeat of the oldest mistake in conversion optimisation: sending qualified traffic to a page that was never designed to receive it. [Pam Didner at Inc.](https://www.inc.com/pam-didner/chatgpt-keeps-sending-customers-to-the-part-of-your-website-nobody-manages/91379788) adds the organisational problem: in most companies internal search has no owner, so nobody watches what happens there. ## What to check on your internal search results page Start with the simplest test: take the ten questions customers ask most often and type them into your site search, one by one, the way someone coming from ChatGPT would. Both articles above propose this exercise; Inc. suggests repeating it every quarter and grading three things per question: the right page is in the top three results, it is up to date, and it actually answers. Then fix what fails, in this order: 1. **No empty pages.** If search finds nothing, show the main categories, the most requested products and a way to contact you. A "0 results" page for someone arriving from an AI recommendation is a lost sale. 2. **Results by intent, not by date.** Someone searching "price", "quote" or "delivery" needs the pricing page, the form or the terms at the top, not the latest blog post that happens to contain that word. 3. **Readable titles and descriptions.** Every result needs a clear title and two lines of useful text: treat the results page as a small landing page. 4. **Synonyms and natural language.** ChatGPT passes whole sentences, not keywords. Your engine has to cope with "how much is shipping to Scotland", not just "shipping". 5. **The parameter stays in the URL.** If the site redirects `/?s=` to the homepage or swaps the parameter, you lose the user's question and never see it in reports. Check the page returns a 200 status and loads fast on mobile. ## Product, pricing and About pages: the other entry points After internal search, ChatGPT visits cluster on the pages that close a decision. In e-commerce 43% land on product pages; in education 52% on course pages; in health 42.1% on the About page, because someone choosing a provider wants to know who is behind it. Previsible draws three practical conclusions: treat product pages as entry points, with structured, comparison-ready data; make pricing machine-readable, because "contact us for a quote" gives the assistant little to work with; and track AI traffic by page type, not site-wide. An example: a travel agency in Edinburgh with a site-wide AI share of 0.3% may find that its pricing page runs at 1% and its individual trip pages at 2%. That is where the time should go, not into the average. For product pages, the feed and structured-data requirements are covered in our article on [ChatGPT Shopping and product feeds](https://geosnap.ai/en/blog/chatgpt-shopping-feed-prodotti-come-preparare-e-commerce); for the About page, the rules of any answer-shaped content apply: who you are, since when, where, what you do, with verifiable numbers. ## How to measure where AI traffic lands in Google Analytics 4 With a source-based comparison and the landing page report. In GA4, build a comparison that includes the sources `chatgpt.com`, `perplexity.ai`, `gemini.google.com`, `copilot.microsoft.com` and `claude.ai` (Previsible uses this list, with app and subdomain variants), then open the Landing page report: you will see at once how much weight falls on URLs with search parameters, product pages and company pages. Also switch on "Site search" in enhanced measurement, so every internal search fires an event carrying the search term: it is the most direct way to read the questions ChatGPT is handing you. If you have connected [Google Analytics to Geosnap](https://geosnap.ai/en/features), the same comparison between monitored prompts and visits received is available without rebuilding the segments. Two caveats. First: Previsible computes penetration as AI sessions divided by total sessions times 100, and recommends doing it per page type; a site at 0.25% overall can have pages at three or five times that. Second: according to [Similarweb's October 2026 update](https://www.similarweb.com/blog/insights/ai-news/ai-search-news-update/), ChatGPT sometimes strips UTM parameters during redirects, so part of the visits end up as direct traffic. If your numbers look low, read how to tell [lost visits from lost labels](https://geosnap.ai/en/blog/traffico-chatgpt-crollato-utm-visite-perse-o-etichette-perse). ## What these numbers do not tell you They do not say how well each landing page converts, and they do not cover Google's AI Overviews. Previsible states this openly: the study measures only referrals from standalone assistants, and the author estimates that AI Overviews drive more traffic than all of them combined, without measuring it. The sample is made of the agency's clients, with a sector mix that does not match any single national market, and the data stops in May 2026. The 58.8% homepage figure is a panel estimate, not Analytics data. So the useful takeaway is not any single percentage but the method: look in your own account at which pages receive AI visits, and fix those. ## Sources - [2026 AI Traffic Report: ChatGPT Wins 92% Share (State of AI Discovery)](https://previsible.com/seo-strategy/ai-traffic-report-july-2026/) (Previsible, 2026-07-06) - [AI Referrals Are Recreating The Oldest Mistake In Conversion Optimization](https://www.searchenginejournal.com/ai-referrals-are-recreating-the-oldest-mistake-in-conversion-optimization/584331/) (Search Engine Journal, 2026-08-05) - [ChatGPT Keeps Sending Customers to the Part of Your Website Nobody Manages](https://www.inc.com/pam-didner/chatgpt-keeps-sending-customers-to-the-part-of-your-website-nobody-manages/91379788) (Inc., 2026-07-26) - [AI Search News: October 2026](https://www.similarweb.com/blog/insights/ai-news/ai-search-news-update/) (Similarweb, 2026-10-21) ## Frequently asked questions ### Why does ChatGPT send people to my site's internal search? Because it has worked out that the answer is on your domain but has no precise, current URL to offer. So it builds a link to your internal search with the user's question as a parameter. Previsible sees this in 28.8% of ChatGPT visits and considers it structural, more frequent in sectors with many similar pages such as SaaS. ### Should I make my internal search results page indexable? No, there is no need: ChatGPT visits are direct clicks by people, not crawls. Keep the noindex if you have one, but check that the page returns a 200 status, keeps the parameter in the URL, does not redirect to the homepage and shows useful results for natural-language questions too. ### How do I see in Google Analytics 4 which pages AI traffic lands on? Build a comparison with the sources chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai, then open the Landing page report. Switch on site search in enhanced measurement to record the search terms. Compute the AI share per page type, not per site: pricing and product pages are often three to five times the average. ### Do these numbers apply to my market? The sample is 166 client sites of one agency, in nine sectors, with no breakdown by country: the percentages do not transfer literally. The method does: look in your own account at which pages receive AI visits and fix those. ChatGPT's internal-search behaviour does not depend on the site's language. ### Are Google's AI Overviews included in the study? No. Previsible measures only referrals from standalone assistants (ChatGPT, Perplexity, Gemini, Copilot, Claude), because AI Overviews cannot be separated from Google organic traffic in GA4. The author estimates they drive more traffic than all the assistants combined, but does not measure it.
Structured data
Schema.org data (JSON-LD) embedded in the page code. It tells the AI explicitly what the page is about and who published it.
BreadcrumbList describes the path through the site
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Home",
"item": "https://geosnap.ai/en/"
},
{
"@type": "ListItem",
"position": 2,
"name": "Blog",
"item": "https://geosnap.ai/en/blog/"
},
{
"@type": "ListItem",
"position": 3,
"name": "Guides",
"item": "https://geosnap.ai/en/blog/#guide"
},
{
"@type": "ListItem",
"position": 4,
"name": "Where does ChatGPT traffic land? 28.8% ends up in your site search",
"item": "https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare"
}
]
}
BlogPosting describes the article: author, dates and sources
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"@id": "https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare#article",
"headline": "Where does ChatGPT traffic land? 28.8% ends up in your site search",
"description": "Previsible's study of 6.77 million sessions: ChatGPT sends 28.8% of visits to internal search, then to product and About pages. What to check and how to measure it.",
"inLanguage": "en",
"url": "https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare",
"mainEntityOfPage": "https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare",
"image": "https://geosnap.ai/img/blog/p2026-11-03-dove-atterra-traffico-chatgpt-en.webp",
"articleSection": "Guides",
"wordCount": 1331,
"author": {
"@type": "Person",
"name": "Rinald Sefa",
"jobTitle": "CMO",
"url": "https://geosnap.ai/en/about",
"worksFor": {
"@id": "https://geosnap.ai/#organization"
}
},
"publisher": {
"@id": "https://geosnap.ai/#organization"
},
"isAccessibleForFree": true,
"abstract": "Visits from ChatGPT do not arrive on blog posts: according to Previsible (6.77 million sessions, 166 sites, November 2024 to May 2026) 28.8% land on the internal search results page, then on product pages (43% in e-commerce), course pages (52%) and About pages (42.1% in health). It happens because the assistant trusts the domain but does not know which page to pick. The fix is to test your site search with the ten most frequent questions, remove empty results pages, push decision pages to the top and measure AI traffic by page type in GA4.",
"datePublished": "2026-11-03",
"dateModified": "2026-11-03",
"citation": [
"https://previsible.com/seo-strategy/ai-traffic-report-july-2026/",
"https://www.searchenginejournal.com/ai-referrals-are-recreating-the-oldest-mistake-in-conversion-optimization/584331/",
"https://www.inc.com/pam-didner/chatgpt-keeps-sending-customers-to-the-part-of-your-website-nobody-manages/91379788",
"https://www.similarweb.com/blog/insights/ai-news/ai-search-news-update/"
],
"translationOfWork": {
"@id": "https://geosnap.ai/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare#article"
}
}
FAQPage describes the frequently asked questions
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "https://geosnap.ai/en/blog/dove-atterra-il-traffico-da-chatgpt-28-per-cento-ricerca-interna-cosa-fare#faq",
"mainEntity": [
{
"@type": "Question",
"name": "Why does ChatGPT send people to my site's internal search?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Because it has worked out that the answer is on your domain but has no precise, current URL to offer. So it builds a link to your internal search with the user's question as a parameter. Previsible sees this in 28.8% of ChatGPT visits and considers it structural, more frequent in sectors with many similar pages such as SaaS."
}
},
{
"@type": "Question",
"name": "Should I make my internal search results page indexable?",
"acceptedAnswer": {
"@type": "Answer",
"text": "No, there is no need: ChatGPT visits are direct clicks by people, not crawls. Keep the noindex if you have one, but check that the page returns a 200 status, keeps the parameter in the URL, does not redirect to the homepage and shows useful results for natural-language questions too."
}
},
{
"@type": "Question",
"name": "How do I see in Google Analytics 4 which pages AI traffic lands on?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Build a comparison with the sources chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai, then open the Landing page report. Switch on site search in enhanced measurement to record the search terms. Compute the AI share per page type, not per site: pricing and product pages are often three to five times the average."
}
},
{
"@type": "Question",
"name": "Do these numbers apply to my market?",
"acceptedAnswer": {
"@type": "Answer",
"text": "The sample is 166 client sites of one agency, in nine sectors, with no breakdown by country: the percentages do not transfer literally. The method does: look in your own account at which pages receive AI visits and fix those. ChatGPT's internal-search behaviour does not depend on the site's language."
}
},
{
"@type": "Question",
"name": "Are Google's AI Overviews included in the study?",
"acceptedAnswer": {
"@type": "Answer",
"text": "No. Previsible measures only referrals from standalone assistants (ChatGPT, Perplexity, Gemini, Copilot, Claude), because AI Overviews cannot be separated from Google organic traffic in GA4. The author estimates they drive more traffic than all the assistants combined, but does not measure it."
}
}
]
}
Metadata
The information in the page header, which crawlers and engines read before the content.