---
title: "Are people leaving Google for ChatGPT? What the data says"
description: "How many visits ChatGPT and AI Overviews send, how assistants pick sources, what has no proven effect, and what a small business should do first."
url: "https://sajdoko.com/blog/are-people-leaving-google-for-chatgpt/"
language: "en"
content_signal: "ai-train=no, search=yes, ai-input=yes"
---
# Are people leaving Google for ChatGPT? What the data says

 15 November 2026·12 minutes read·[SEO](https://sajdoko.com/blog/category/wordpress/seo/)     

Business owners now hear two opposite claims in the same week. One says Google is finished and customers ask ChatGPT instead. The other says AI search is hype and nothing needs to change. We read the primary studies and the official documentation from Google, OpenAI, Anthropic, Perplexity and Cloudflare to see what can be supported with data.

 

The short version: people do use AI assistants, AI summaries do reduce clicks, and the visits that assistants send to websites are still a very small share of traffic. Most of what makes a business visible in AI answers is the same work that makes it visible in Google. Several things sold as “AI optimisation” have no measured effect at all.

 

Every number below comes with its source and its limits. Studies measure different things, and most of the confusion in this field comes from mixing them.

 

## Google is still where the searches happen

 

Similarweb estimates that in October 2026 ChatGPT received 6.05 billion website visits, Gemini 2.64 billion, Claude 1.25 billion and Perplexity 88.3 million. Together that is about 10 billion visits. Google.com received 88.44 billion in the same month. These are modelled visits to websites and exclude mobile apps, so they show scale, not search market share. [Similarweb: most visited websites](https://www.similarweb.com/blog/research/market-research/most-visited-websites/)

 Google.com received almost nine times the visits of the four assistants combined. 

Among traditional search engines, Statcounter puts Google at 91.05% worldwide for October 2026 and Bing at 4.80%. [Statcounter: search engine market share](https://gs.statcounter.com/search-engine-market-share/worldwide)

 

A visit to an assistant is also not a search for a business. People use these tools to write, translate, summarise and code. For a small business, the practical conclusion is that Google remains the main source of search visits, and AI assistants are a second channel that is growing next to it.

 

## When an AI summary appears, fewer people click

 

Three studies are quoted most often. They measure three different things and should not be merged into one figure.

  

| Study | What it measured | Result |
| :--- | :--- | :--- |
| Pew Research Center | 900 US adults, 68,879 Google searches, March 2025 | A traditional result was clicked in 8% of visits with an AI summary and 15% without one. A link inside the summary was clicked in 1% of visits. |
| SparkToro and Similarweb | All US Google searches, January to April 2026, web panel | 68.01% of searches ended without a click. This covers every search, with or without an AI summary. |
| Seer Interactive | 53 brands, about 5.47 million tracked queries | Organic click-through rate was 2.36% on queries with an AI Overview and 3.82% without, in February 2026. |

  

Sources: [Pew Research Center](https://www.pewresearch.org/short-reads/2025/07/21/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/), [SparkToro](https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/), [Seer Interactive](https://www.seerinteractive.com/insights/aio-impact-on-google-ctr-2026-update).

 Both studies point the same way, but they measure different things. 

The direction is the same in all three: a summary at the top of the page takes clicks from the results below it. There is no single “zero-click rate” you can apply to your own site. What you can do is check in Search Console which of your queries lost clicks while impressions held steady. Those are usually informational questions. Searches by someone ready to book, buy or ask for a quote are affected less, because a summary cannot complete the booking.

 

## AI assistants send few visits so far

 

Semrush looked at more than 50,000 websites in 17 industries through 2025. Visits from AI tools grew 66% during the year and still made up less than 0.15% of all visits in the dataset. Organic search was far larger. [Semrush: traffic channel study](https://www.semrush.com/blog/traffic-channel-mix-study/)

 

Do those visitors convert better? The studies disagree, and the reason is who was measured:

 

- Amsive compared 54 sites over six months. AI visits converted at 4.87% and organic visits at 4.60%. The difference was not statistically significant. [Amsive](https://www.amsive.com/insights/seo/does-llm-traffic-convert-better-than-organic-a-new-data-backed-study/)
- Marie Haynes analysed 100 sites and found ChatGPT visits converting at 13.8% against 3.02% for Google organic. In the same data, Google still produced 96.5% of all measured conversions. [Marie Haynes](https://www.mariehaynes.com/geo-is-a-distraction/)
- Adobe measured US retailers in September 2026 and reported 90% higher purchase conversion and 83% higher revenue per visit from AI referrals, compared with all other channels combined. [Adobe](https://business.adobe.com/blog/ai-traffic-higher-revenue-per-visit-retail-holiday)

 The channel is small, and the conversion studies do not agree. 

None of these gives a multiplier you can borrow. A clinic, a tour operator or a service company should count its own enquiries by source and see what they are worth.

 

One more finding is useful for planning. Ahrefs reported that over 80% of AI referrals to its own site landed on the homepage, product pages and free tools, not on blog articles. It is one software company’s data, so treat it as an example. [Ahrefs](https://ahrefs.com/blog/ai-search-traffic-by-page-type-ahrefs/) The lesson still holds for a small site: an article may be what the assistant reads, while the visitor arrives on the homepage or a service page. Both need to be good, and they need to link to each other.

 

## How assistants choose what to cite

 

Ranking in Google’s top ten for a query does not mean an assistant will cite you for it. Ahrefs compared the exact pages cited by assistants with the pages ranking in Google for the same prompt:

  

| Where the answer appears | Cited pages that are also in Google’s top 10 |
| :--- | :--- |
| ChatGPT | about 8% |
| Gemini | about 8% |
| Perplexity | 28.6% |
| Google AI Overviews | 37.1% |
| Claude | no comparable figure published |

  

Sources: [Ahrefs, 15,000 prompts](https://ahrefs.com/blog/ai-search-overlap/) and [Ahrefs, 4 million AI Overview citations](https://ahrefs.com/blog/ai-overview-citations-top-10/). For AI Overviews, a further 26.2% of cited pages ranked between positions 11 and 100, and 36.7% were outside the top 100.

 Most pages that assistants cite are not in Google’s top ten for the same question. 

Google explains part of this itself. AI Overviews and AI Mode can split one question into several related searches, then collect pages for each. Google calls it query fan-out. A page can be cited because it answers one of those related questions well, even if it does not rank for the original wording. [Google: AI features and your website](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)

 An illustrative example: the pricing page is cited through a related search. 

For a business site this argues for complete service pages. A page that states what the service includes, who it suits, what it costs or what the price depends on, how long it takes and what happens next can match many related questions. Ten thin pages that repeat one keyword cannot.

 

## What goes together with being mentioned

 

Ahrefs studied 10,000 brands across 260 million AI prompts, published on 3 November 2026. Brands that assistants mentioned more often were also mentioned more often elsewhere on the web: in Reddit threads that rank in search (correlation 0.75), on other websites (0.70), in the press (0.67), on YouTube (0.64) and on TikTok (0.60). [Ahrefs: mention types](https://ahrefs.com/blog/mention-types-ai-visibility/)

 

An earlier Ahrefs study of 75,000 brands found a similar link with web mentions and much weaker links with branded search volume (0.352) and its own Domain Rating score (0.266). That sample only included established brands, so it says little about a new or very small business. [Ahrefs: brand visibility factors](https://ahrefs.com/blog/ai-brand-visibility-correlations/)

 

These are correlations. They do not show that one more mention produces one more recommendation, and nobody outside the AI companies knows the weights. Buying mentions or posting fake reviews on that basis would be a mistake. Real listings in directories and professional associations, honest reviews and articles written by others are worth having anyway, because customers read them too.

 

## Search crawlers and training crawlers are separate

 

Each AI company runs more than one crawler, and they do different jobs. This is documented, production behaviour.

  

| Company | Crawlers for search and live answers | Crawler or control for model training |
| :--- | :--- | :--- |
| OpenAI | OAI-SearchBot, ChatGPT-User | GPTBot |
| Anthropic | Claude-SearchBot, Claude-User | ClaudeBot |
| Perplexity | PerplexityBot, Perplexity-User | see its current documentation |
| Google | Googlebot | Google-Extended (a control, not a separate crawler) |

  

Sources: [OpenAI](https://developers.openai.com/api/docs/bots), [Anthropic](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler), [Perplexity](https://docs.perplexity.ai/docs/resources/perplexity-crawlers), [Google](https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers).

 Blocking a search crawler removes you from answers. Blocking a training crawler does not. 

This means you can let ChatGPT find and cite your pages while refusing their use for training:

 

```
User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /
```

 

This is an illustration. Keep your other rules and test the live file afterwards. OpenAI says a change can take about 24 hours to show in search. Google-Extended does not affect whether you appear in Google Search or AI Overviews. Blocking training crawlers is a decision about how your content is used. It does not improve or reduce your visibility in answers.

 

The more common problem is accidental blocking. A security plugin, a firewall rule or a CDN challenge page can return an error to a legitimate search crawler while the site looks fine in a browser. No study tells us how many small business sites have this problem, which is a reason to check your own.

 

## What has no proven effect

 

Several newer mechanisms are discussed as if they bring recommendations. The documentation says otherwise.

  

| Mechanism | Status in 2026 | What it is for |
| :--- | :--- | :--- |
| llms.txt | Optional convention. Google states that Search ignores it. | A summary of the site for tools that choose to read it |
| Content-Signal | A Cloudflare format for stating content-use preferences | Policy on search, AI input and training. Not a ranking input. |
| AIPREF training preference | An individual IETF draft, not a finished standard | Stating a training preference in a standard way |
| WebMCP | A Chrome origin trial | Letting browser agents use a site’s forms and tools |
| Generic JSON-LD | Established for rich results | Describing the business accurately. Not a special AI requirement. |

  

Sources: [Google](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide), [Cloudflare](https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/), [IETF draft](https://datatracker.ietf.org/doc/draft-kelley-aipref-training-preference-signal/), [Chrome](https://developer.chrome.com/blog/ai-webmcp-origin-trial).

 

We publish several of these on our own site, because we build tools that agents can use and we want to know how the standards behave in practice. We do not present them as a way to be recommended, and you should be careful with anyone who does. A protocol can be useful without changing who gets mentioned.

 

## What to do, in order

 

The evidence supports a plain sequence: make the site reachable and clear, make the business consistent everywhere, then build reputation, then monitor.

 

**First two weeks**

 

1. Check that your main pages are indexed, return a normal response and are not blocked by robots.txt, a noindex tag or a firewall. Search Console’s URL inspection shows what Google sees.
2. Put the facts in plain HTML that loads without interaction: business name, services, areas served, qualifications, prices or what the price depends on, and how to get in touch.
3. Set up enquiry and purchase events in Google Analytics, and group visits from AI assistants as their own source so you can see them.

 

**First month**

 

1. Make the business name, address and phone number identical on the website, Google Business Profile and directory listings.
2. Write one complete page per service and per location, with the details a buyer needs to decide and a clear way to book or ask.
3. Add accurate structured data for the business, its services and breadcrumbs where it applies.
4. Ask real customers for reviews. Publish case studies, credentials and your own photos.

 

**From the second month**

 

1. Answer the questions buyers ask before they call: cost, suitability, process, alternatives.
2. Earn mentions through local and industry contributions, memberships and genuine publicity.
3. Once a month, ask each assistant the same fixed set of questions in your language and note which businesses it recommends, which pages it cites and what it gets wrong about you.

 

What a complete page contains depends on the business. A clinic needs practitioner qualifications, treatment pages with who each treatment suits, booking instructions and prices. A tour operator needs each tour’s route, duration, inclusions, season, and booking and refund terms. A service company needs its specialisation, process, scope of delivery, case studies and how a quote works.

 

## How to measure it

 

Google added a generative AI performance report to Search Console in June 2026 and finished the worldwide rollout by the end of August. It shows impressions and the pages shown in AI features. [Google Search Central](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports) It does not tell you whether anyone became a customer, so pair it with your own numbers:

 

- **Shown:** impressions in the Search Console report.
- **Mentioned:** how often your fixed set of questions returns your business, by assistant and date.
- **Cited:** which of your pages are linked, and whether those links open.
- **Visited:** sessions from AI sources in Analytics.
- **Converted:** enquiries, calls and bookings from those sessions, and what they were worth.
- **Accurate:** whether the assistant states your prices, address and services correctly.

 

Being mentioned, being cited and receiving a visit are three separate results. A report that mixes them will mislead you.

 

## Where this fits our work

 

Our [free AI-readiness check](https://sajdoko.com/ai-ready/) covers the first step of the list above: whether assistants and search crawlers can reach your pages, whether the basic business facts are readable, and which crawlers your robots.txt allows or blocks. It labels the emerging standards as such and scores them low. No tool, including ours, can guarantee a recommendation.

 

The rest is ordinary, careful work: clear service pages, consistent business details and measurement, which is what our [SEO service](https://sajdoko.com/services/seo/) covers. If you want to know where your site stands, run the check and send us the report.

 

## Frequently asked questions

  Has AI search replaced Google? 

No. In October 2026 Google.com received about 88 billion website visits against about 10 billion for ChatGPT, Gemini, Claude and Perplexity together, and much of the assistant use is not search. AI assistants are an additional channel that is growing.

   Do visitors from ChatGPT convert better than visitors from Google? 

It depends on the site. One study found no significant difference, another found a large one on a selected sample, and a retail study found higher conversion against all other channels combined. Measure your own enquiries by source before assuming anything.

   Do I need an llms.txt file to appear in AI answers? 

No. Google states that Search ignores it, and no study shows that it increases recommendations in other assistants. It does no harm, and it comes after indexing, clear service pages and consistent business details.

   Should I block AI crawlers? 

Decide separately for search and for training. Blocking a search crawler such as OAI-SearchBot removes you from that assistant’s answers. Blocking a training crawler such as GPTBot only refuses the use of your content for training and does not affect whether you are cited.

   How do I know if assistants recommend my business? 

Write down ten to twenty questions a customer would ask, in the language they would use. Ask each assistant the same questions once a month and record which businesses it names, which pages it links and any wrong facts about you. The Search Console report adds impressions for Google’s AI features.

   

- [SEO](https://sajdoko.com/blog/category/wordpress/seo/)

   

## Have a project in mind?

 

Write to us and we come back with a concrete estimate within one working day.

  [Get a quote →](https://sajdoko.com/contact/)     

## Leave a comment [Cancel](https://sajdoko.com/blog/are-people-leaving-google-for-chatgpt/?format=markdown#respond)

      

## Recent Posts

- [Are people leaving Google for ChatGPT? What the data says](https://sajdoko.com/blog/are-people-leaving-google-for-chatgpt/)
- [Website redesign SEO checklist: what to preserve before launch](https://sajdoko.com/blog/website-redesign-seo-checklist/)
- [Fleet management software: what do you need beyond GPS?](https://sajdoko.com/blog/fleet-management-software-requirements-beyond-gps/)
- [Custom CRM Development vs Generic SaaS: The ROI Framework for Growing Enterprises](https://sajdoko.com/blog/custom-crm-development-vs-generic-saas/)
- [Web and Software Development Costs in Albania: The International Buyer’s Strategic Guide](https://sajdoko.com/blog/website-costs-in-albania-international-perspective/)

## Recent Comments

1. Alvaro on [Optimizing and Extending the Livewire Product Filter – Part 4](https://sajdoko.com/blog/optimizing-and-extending-the-livewire-product-filter/#comment-4)
2. [sajdoko](https://sajdoko.com) on [How to Set Up WordPress on localhost Using XAMPP: A Step-by-Step Guide](https://sajdoko.com/blog/how-to-install-wordpress-on-localhost-using-xampp-step-by-step-guide/#comment-3)
3. Roshni on [How to Set Up WordPress on localhost Using XAMPP: A Step-by-Step Guide](https://sajdoko.com/blog/how-to-install-wordpress-on-localhost-using-xampp-step-by-step-guide/#comment-2)