What changes between showing up on Google and showing up in ChatGPT

Ask ChatGPT a question and, in most answers, it doesn’t run any searches of its own. Ahrefs reviewed 66.6 million answers that the free version of ChatGPT gave in August and September 2026. Only 14.4% of them included searches that ChatGPT wrote on its own. That’s one in seven.

That number shows the first difference from Google. On Google, every search produces a results page, the famous SERP. SEO is about winning a spot on that list. In ChatGPT there are two ways to show up. One is that the model already knows about you when it answers from memory. The other is that it goes out to search, finds your page and picks it to cite.

Google ranks a list. ChatGPT picks who to cite

When ChatGPT searches, it almost never uses your question as is. It rewrites it and splits it into several smaller searches. SEOs call this query fan-out. I’ll call those smaller searches fan-out queries.

AirOps studied that process with 15,000 prompts and published the results in March 2026. When ChatGPT searched, it ran two or more fan-out queries 89.6% of the time. In total it read 548,534 pages and cited about 15% of them.

That’s the second difference. On Google, if your page is on the list, it gets an impression even if nobody clicks. In ChatGPT, the model can read your page and still leave it out of the answer.

According to the same study, 95% of fan-out queries had zero search volume in the metrics that keyword tools use. Keyword research tools won’t show them to you.

Does ranking first on Google help you show up in ChatGPT?

It helps a lot. AirOps compared the pages ChatGPT read with Google’s top 20, both for the original prompt and for each fan-out query. ChatGPT cited 43.2% of the pages that ranked first for at least one of those searches. Those pages were 3.5 times more likely to be cited than pages outside the top 20.

The result changes depending on which search you measure against. In 2025, Ahrefs compared citations from ChatGPT, Gemini and Copilot with Google’s top 10 for the original prompt, across 15,000 cases. Only 12% of the URLs those AI tools cited were in that top 10.

Fan-out queries explain the gap between the two numbers. AirOps found that, among cited pages that made Google’s top 20 for at least one search, 32.9% ranked only for a ChatGPT fan-out query. For the original prompt, those pages were outside the top 20.

Classic SEO gets you into the set of pages ChatGPT reads. To get cited, your page also has to answer the fan-out queries for its topic. Ahrefs offers a hint at what they look like. Among the words ChatGPT added to its searches most often in September 2026 were “2026”, “official” and “price”. If your page doesn’t give the year, the price or the official figures, another page will.

What other sites say about you

In SEO, a brand mention without a link used to be worth little. In AI answers, those mentions seem to carry a lot more weight.

Ahrefs analyzed 75,000 brands in ChatGPT and in Google’s AI features, AI Mode and AI Overviews, and published the study in December 2025. It correlated how often each AI named a brand with metrics like links, traffic and mentions. The strongest correlation was with the number of brand mentions in YouTube videos, around 0.74 on all three platforms. Mentions on other websites came next, between 0.66 and 0.71. In ChatGPT, Domain Rating (the Ahrefs metric that estimates the strength of a site’s links) showed a weak correlation of 0.266.

Keep in mind these are correlations. They show that two things move together, without proving that one causes the other. They still help you decide where to put your effort. Today, getting named in a YouTube video in your industry looks more valuable than one more link.

The technical part: one more bot to let in

ChatGPT uses its own bots to read the web. To show up in its searches, your site has to let OAI-SearchBot in. According to OpenAI’s documentation, if you block it in robots.txt, your pages are left out of those answers and can only appear as a navigational link. GPTBot is a different bot, the one that collects content to train the models, and blocking it is a separate decision.

Before anything else, open your robots.txt and check that OAI-SearchBot can get in.

Google Search, by contrast, needs little that’s new. In its guide to generative AI features, published on May 15, 2026, Google says the usual SEO work is enough to show up there. You don’t need files like llms.txt or special markup. For Google, optimizing for those features is SEO. That guide applies to Google Search. For ChatGPT, OpenAI’s documentation is what counts.

How do you measure each one?

On Google you have Search Console, with impressions, clicks and position for each query. OpenAI doesn’t offer anything like it for ChatGPT yet. In the meantime, you can track visits from chatgpt.com in your analytics. You can also check the same set of prompts every month with a paid monitoring tool.

SEO consultant Marie Haynes analyzed 12 months of Google Analytics data from 100 sites, and MarTech published the results in October 2026. ChatGPT sent the equivalent of 0.62% of the visits that came from organic Google search. Across the 55 sites that tracked conversions, people who arrived from ChatGPT converted more: 13.8% versus 3.02% for organic Google. Organic Google still brought in 27 times more conversions than all AI assistants combined, ChatGPT included.

My bet is to do both with the same groundwork, without dropping Google for ChatGPT.

And yes, this post is also written so ChatGPT will cite it. I’ll let you know if it picks me.