ChatGPT SEO: How to Show Up in ChatGPT's Answers

ChatGPT SEO means making your pages findable, readable and citable at the moment ChatGPT searches the web to answer a question. In practice: let its search crawler, OAI-SearchBot, in, be properly indexed, publish pages that give a clear answer to a real question, and exist on the sites ChatGPT consults. None of this can be bought.

It is the most frequent request I have received over the past year. A business owner asked ChatGPT about their industry, a competitor came up, they did not. Here is how ChatGPT chooses its sources, the technical settings to check, the seven steps I apply to get into its answers and how to measure the results.

What is ChatGPT SEO?

ChatGPT SEO, also called SEO for ChatGPT, covers the actions that increase your chances of appearing in ChatGPT's answers: being cited as a source, with a link to your page, or being recommended by name. It is generative engine optimization (GEO) applied to one specific engine.

ChatGPT can talk about you in two ways, and you do not work on them the same way.

ChatGPT SEO is mostly played out in the second column. The model's memory reflects years of online presence; web search is something you can work on this week. It has been open to all users, even without an account, since February 2025.

How ChatGPT chooses its sources

Understanding the mechanism keeps you from spending your energy in the wrong place. The process happens in four steps, and each one is a filter.

1. The search is triggered, or not

ChatGPT does not systematically search the web. On a general, stable question, it answers from memory, based on its training. In that case, no source is cited and your optimization work has no effect.

A search is triggered when the question calls for fresh or verifiable information: a price, a date, news, a recommendation for a provider, a product comparison. Those are the questions worth fighting for.

2. The index: OAI-SearchBot and third-party engines

When a search is triggered, ChatGPT rephrases the question into several queries and sends them to search indexes. It does not browse the web at random. OpenAI says it relies on third-party search providers and on its own crawler, OAI-SearchBot, which crawls pages for ChatGPT search.

The direct consequence: a page missing from those indexes, or off-limits to OAI-SearchBot, does not exist for ChatGPT. Traditional SEO remains the prerequisite. Back in 2023, ChatGPT's web browsing was actually called "Browse with Bing". OpenAI no longer details its providers, but checking your indexing in Bing is still the simplest precaution.

3. Extracting the passage

The system retrieves a handful of pages and looks, in each one, for the passage that answers the query. This is where most sites lose, and it is the filter you have the most power over.

An extractable passage makes sense on its own. If it starts with "That is why", if it refers back to a previous paragraph, or if the answer comes after six paragraphs of context, it will not be picked.

4. The final selection

Among the passages it has kept, ChatGPT retains a few to build its answer. For answers of equal quality, it favors sources that look reliable: well-known sites, identified authors, information corroborated elsewhere.

That is where your brand's reputation outside your own site comes in, and it is the slowest lever to build.

OAI-SearchBot, GPTBot, ChatGPT-User: which crawlers should you allow?

OpenAI publishes the list of its crawlers and their roles in its developer documentation. They are independent: each one is configured separately in robots.txt.

To be visible on ChatGPT, the minimum is to allow OAI-SearchBot. Here is the robots.txt I recommend to a business that wants to be cited and known by the models:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Allow: /

Sitemap: https://www.your-site.com/sitemap.xml

If you do not want your content used for training, replace Allow: / with Disallow: / under GPTBot only. You stay eligible for search. OpenAI says a robots.txt change takes about 24 hours to be picked up.

What about the llms.txt file? OpenAI does not mention it in its crawler documentation, and it does nothing to replace OAI-SearchBot's access. I have covered its real value in my article on the llms.txt file.

How to get cited by ChatGPT: the 7 steps

Here is the order I work in. The first two steps are technical and quick, the next five do most of the work.

Step 1: open access to OAI-SearchBot

Open your-site.com/robots.txt and look for a Disallow rule aimed at OAI-SearchBot, or an overly restrictive User-agent: *. Then check your CDN's AI crawler setting. To confirm, search for "OAI-SearchBot" in your server logs: regular visits show the crawler is getting through.

Step 2: check your indexing, Bing included

Check your key pages in Google Search Console and in Bing Webmaster Tools, which lets you import your sites from Search Console in a few clicks. Submit your sitemap to both. Finally, make sure the text is in the HTML that is served: content that only appears after scripts run may never be read.

Step 3: target the questions that trigger a search

Make a list of the questions your customers ask before buying: prices, lead times, comparisons, "best X for Y", regulatory requirements with deadlines. Keep the ones that call for fresh information. Each one deserves a page, or at least a clearly titled section.

Step 4: answer in the first two sentences

If you only do one thing, do this. Take your ten most important pages and rewrite their first paragraph so it contains the complete answer to the question the page covers. The first sentence restates the question as a statement, the next one gives the key point, the rest of the page expands. It is the structure of a wire story: you can cut it anywhere without losing the meaning.

Step 5: add dates and figures, and split into self-contained blocks

A verifiable claim is preferred to a general one. "OpenAI says a robots.txt change takes about 24 hours to be picked up" will get reused; "processing times vary" never will. Dated facts have a valuable side effect: they are what pushes ChatGPT to go and search.

Each section must also read on its own: a heading that contains the question, short paragraphs on one idea, lists when the information is a list, tables when it is a comparison. The test is instant: take a paragraph out of the page and read it. If it stands on its own, it is extractable. That is the practical core of AEO (answer engine optimization), which I cover in a separate guide.

Step 6: get into the sources ChatGPT already cites

"Which provider would you recommend for this": that is the question that is worth money. For this type of request, ChatGPT rarely goes to the companies' own websites. It goes where someone has already done the sorting: comparison sites, business directories, "best X in Y" articles, forums, online reviews. To get recommended, your presence elsewhere matters more than your own website.

1. Identify the cited sources. Ask ten recommendation questions from your industry and note the sites ChatGPT uses to answer.
2. Get listed there. Many accept a listing, free or cheap: business directories, trade associations, platforms specialized in your field.
3. Grow your reviews. For local businesses, they weigh heavily in recommendations, just as in Google's local pack.

This step often produces results faster than content production, because it does not depend on your own site being recrawled.

Step 7: make your brand identifiable

ChatGPT cross-checks information. The same name, the same address and the same description everywhere, an "about" page that says who you are, identified authors on your articles and consistent Organization markup make the entity stronger. One inconsistency between two directories is enough to weaken it.

The levers, ranked by effectiveness

For steps 4 and 5, test a page before rewriting it. Paste its text and the target question: the tool applies the criteria an answer engine uses to extract a passage, and tells you what is missing.

Visibility on ChatGPT: what changes depending on your business

The general method is the same for everyone, but the priority lever depends on what you sell. Here is how I adjust case by case.

In four cases out of five, the decisive lever lies outside your website. That is the main change compared with traditional SEO, where optimizing your own pages did most of the work. For an online store, OpenAI also publishes a product feed specification that powers product discovery in ChatGPT: you work on it like a Google Shopping feed.

How to measure your ChatGPT SEO

Three measurements complement each other: citations, traffic and crawler visits.

Citations: a question panel

Measuring citations takes a protocol, otherwise the numbers mean nothing.

1. In a private window, logged out. Otherwise ChatGPT cites you because it knows you, and you are measuring your own history.
2. With stable wording. A rephrased question gets a different answer: you could no longer compare from one week to the next.
3. Several times, over several days. A single answer is an anecdote, not data.

For each question, note three things: are you cited, which competitors are, which sources are linked. The last column is the most useful: it gives you the exact list of sites where you need to be. The full method, with how to build the panel, is in my guide to auditing your AI visibility. If you would rather delegate, that is what my AI visibility audit is for.

Traffic: utm_source=chatgpt.com in Google Analytics

Links clicked in ChatGPT's answers usually arrive with the utm_source=chatgpt.com parameter. In Google Analytics 4, open the "Traffic acquisition" report, display the "Session source" dimension and look for chatgpt.com. Some visits arrive without this parameter and get lost in direct traffic: the figure you get is a floor.

Crawler visits: your server logs

Count OAI-SearchBot's visits in your logs, page by page. A key page that is never visited will never be cited. Since anyone can spoof a crawler's name, cross-check the IP addresses against the list OpenAI publishes for OAI-SearchBot.

The mistakes that keep you from being cited by ChatGPT

• Blocking OAI-SearchBot when you meant to block training. GPTBot is the one used for training. I still regularly find robots.txt files that shut everything out, often without anyone having decided to.
• Letting the CDN filter out AI crawlers. robots.txt says yes, the firewall says no, and the crawler does not get through.
• Burying the answer. Three paragraphs of introduction before getting to the point are enough to disqualify the page.
• Writing creative headings. A pun means nothing to a machine looking for a match with a question.
• Targeting questions that do not trigger a search. ChatGPT answers them from memory, without citing anyone.
• Confusing citations with traffic. Being cited does not guarantee a click. They are two separate goals, to be measured separately.

The four-week plan

Here is the exact order I follow with a client starting from scratch.

1. Week 1: measure and unblock. Build a panel of twenty questions, record the starting point, check robots.txt, the CDN and indexing.
2. Week 2: rewrite page openings. The ten key pages, with the intro redone to answer in two sentences.
3. Week 3: date and structure. Add dated facts, split into blocks, add a marked-up FAQ where it makes sense.
4. Week 4: go outside. Request listings in the comparisons and directories spotted during the measurement.

Then I wait, and measure again after a month. The first movements usually appear between four and eight weeks, the time it takes for crawlers to come back and for indexes to update.

What money cannot buy

There is no way to buy a citation in an answer, no listing in an official ChatGPT directory, no hidden setting. Since August 24, 2026, ChatGPT has indeed shown ads in France, below the answers on the Free and Go plans, but OpenAI states that they are labeled as such, kept separate from the answer, and do not influence it.

So be wary of any offer that guarantees a presence in ChatGPT's answers. Answers vary from one user to another and from one day to the next: nobody can guarantee what they do not control. What does exist is publishing work, the kind I do in my GEO consulting projects: making your pages useful, precise and verifiable, and making sure others talk about you.

Written by

Anthony Courtin

SEO, GEO & AI automation consultant

I help small and mid-sized businesses and independents with their visibility (Google, ChatGPT, Perplexity) and with how they actually use AI at work. I also work on-site with companies to train teams and rethink processes. Everything I write here, I use on client projects.

Follow me on LinkedIn My YouTube channel (in French)