On this page
- ChatGPT often searches before it writes
- Check ChatGPT's crawler can actually reach you
- What Google says, in Google's own words
- For a local business, your website usually isn't the only source
- Write pages worth quoting, starting with things only you know
- How to check whether ChatGPT names you
- How to measure it, and the bit nobody mentions
- The honest summary
- FAQs
A few of our clients have asked some version of the same question this year. Their customers have stopped typing "electrician near me" into Google and started asking ChatGPT who they should call. So how do you get ChatGPT to say your name?
First, a correction to the question. There is no ChatGPT ranking. No numbered list, no position three, nothing you can track week to week the way you can with Google. Ask the same question twice and you can get two different sets of businesses, because it ran two different searches. What you're actually trying to do is be one of the sources it finds and repeats. So when this article says "ranked in ChatGPT", read it as "reliably showing up in the answers", because that's the honest version.
Second, there's now a whole industry selling answers to this. Generative engine optimisation. Answer engine optimisation. AI search optimisation. AI SEO packages. Special files you're told to upload so the robots can read your site properly.
Most of it is nonsense, and you don't have to take my word for that. OpenAI and Google both publish documentation on how this works. Below I've separated three things that usually get mixed together: what the documentation actually says, what's reasonable SEO inference, and what we've seen on our own site. They're labelled, so you can weigh them differently.
ChatGPT often searches before it writes
This is the part that clears up almost everything else.
What the documentation says. ChatGPT doesn't search every time. OpenAI's position is that it will search the web when a question might benefit from current information, and that a single question can turn into several rewritten queries across more than one search provider. For "who should I get to rewire a house in Mansfield", that's exactly the sort of question that triggers a search. For "what is a consumer unit", it may well answer from what the model already knows.
So the framing that matters is narrower than the one you'll usually read: for current and local questions, ChatGPT can search before it answers, and if you want to be cited in those answers, retrieval comes first.
Two rounds, not one. The retrieval round, then the writing round.
Almost everything sold as "AI optimisation" is aimed at the writing round: phrasing tricks, special files, magic formatting. But you can't be summarised out of a set of pages you weren't retrieved into. If you don't win the first round, the second round never involves you.
That's why the practical advice below looks so much like ordinary SEO. Being findable is the strategy.
It also settles a question people ask a lot: whether you need to get your business "into ChatGPT's training data". You don't, and it's the wrong target. Search visibility and model training are controlled separately and serve different purposes, which is what the next section is about. The answer you're trying to appear in was assembled from a live search moments before you read it, not recalled from training.
Check ChatGPT's crawler can actually reach you
Here's the one genuinely new job, and it takes about ten minutes.
The major AI search products publish crawlers or user agents you can identify, and your robots.txt file decides which ones get in. Most site owners have never looked. Some have been handed a "privacy hardening" or "bot protection" setting by a hosting company or a plugin that quietly blocked the lot.
The names matter, because they do different jobs. Blocking the one that trains models is a legitimate choice. Blocking the one that puts you in results is self-harm.
| Crawler | Run by | What it does | If you block it |
|---|---|---|---|
| OAI-SearchBot | OpenAI | Surfaces sites in ChatGPT's search features | You won't be shown in ChatGPT search answers, though you can still appear as a navigational link |
| GPTBot | OpenAI | Crawls content that may be used to train OpenAI's models | Your content isn't used for training. ChatGPT search is unaffected |
| ChatGPT-User | OpenAI | Fetches a page when a user asks ChatGPT to look at it | Limited effect, as user-initiated fetches may not follow robots.txt |
| Googlebot | Crawls for Google Search | Google can't crawl the page, which effectively destroys its useful Search and AI visibility | |
| Google-Extended | Controls use of your content for Gemini and other generative products | Gemini is affected. Your Search and AI Overviews eligibility is not | |
| PerplexityBot | Perplexity | Surfaces and links sites in Perplexity results | You stop appearing as a Perplexity source |
| Perplexity-User | Perplexity | Fetches pages when a user asks it to | Limited effect, as it generally ignores robots.txt |
Sources for that table are OpenAI's crawler documentation and Perplexity's bot documentation.
The distinction worth holding onto: GPTBot is about training. Google-Extended controls training and grounding in some of Google's Gemini products. Neither is the control for appearing in ordinary Google Search. OAI-SearchBot and PerplexityBot are the ones about being found. The controls are separate and serve different purposes, though OpenAI notes it may reuse the result of a single crawl for both when you allow both bots. Plenty of businesses have blocked the lot to keep their content out of AI training, then wondered why they never get cited. Those are different goals and you have to decide which you care about.
One precision on the Googlebot row, because it gets overstated. Blocking Googlebot in robots.txt stops Google crawling the page. A URL blocked that way can still stay in the index if Google finds it through links elsewhere, but it can't be crawled properly and tends to show up without a useful snippet. Since Google's stated bar for its AI features is that a page is indexed and eligible to show a snippet, blocking Googlebot is functionally self-harm even though "you disappear entirely" is too strong.
To check your own site, type your domain followed by /robots.txt into a browser. So yourbusiness.co.uk/robots.txt. If you see Disallow: / under any of the bots above, that's your answer.
If there's no robots.txt at all, that does not mean nothing is blocked. It only means robots.txt isn't doing the blocking. A CDN, firewall, bot-management service, login wall or server rule can refuse these crawlers before robots.txt is ever consulted, and aggressive bot protection is a common default. Cloudflare and similar services are the usual culprits. OpenAI's own recommendation is to allow OAI-SearchBot in robots.txt and allow its published IP ranges at the network layer, which is the bit most people miss. If your site sits behind any kind of bot protection, that's where to look next.
What Google says, in Google's own words
ChatGPT isn't the only place this happens, and Google has been the most direct about how it works. Its guidance is worth reading even if ChatGPT is your main concern, because it's the clearest public statement anyone has made.
Google publishes a page on how sites appear in AI Overviews and AI Mode. It is refreshingly blunt.
On eligibility, it says a page must be indexed and eligible to be shown in Google Search with a snippet. That's the bar. If your page is indexed and can show a normal search snippet, it can be used in an AI Overview.
On optimisation, it says there are no additional requirements to appear in AI Overviews or AI Mode, and no other special optimisations necessary.
There is now a separate control, and it landed in the UK first
This changed in June 2026, and plenty of articles still have the old version.
It used to be true that AI was simply part of Search with no separate switch. Google now provides a Search generative AI control in Search Console, under Settings. It gives you a choice for the property: include your content in Search generative AI features, or exclude it. Inclusion is the default, and child properties inherit the parent's setting unless you override it.
Two things worth knowing. It's a grounding control, not a ranking one. Google is explicit that it "isn't used as a ranking or inclusion signal affecting other parts of Search", so excluding yourself doesn't damage normal rankings. And it's still rolling out to a subset of owners rather than being universally available, so it may not be in your account yet.
The UK detail matters if you're reading this from Britain. The control rolled out to UK properties first, in June 2026, before expanding to other regions. So UK site owners are more likely to have it than most.
For almost every service business the right setting is the default. Being included is how you get recommended. But it's worth knowing the switch exists, and worth checking nobody has flipped it on your behalf.
The llms.txt question, settled
You may have been told to add an llms.txt file, a plain-text summary of your site supposedly written for language models. It's been sold hard as an AI SEO essential.
Google's own AI features documentation settles it in one line: "You don't need to create new machine readable files, AI text files, or markup to appear in these features." The same page adds that there's "no special schema.org structured data that you need to add" either.
Adding one is harmless. Paying someone for it is not. If it's on your quote as a line item, ask what else on that quote was chosen the same way.
For a local business, your website usually isn't the only source
This is the part that surprises people, and it matters most if you're a plumber, electrician, roofer or landscaper.
What the documentation says. OpenAI describes ChatGPT Search as using web search along with trusted third-party providers for things like nearby business listings. That's the extent of what's actually published. It does not say ChatGPT reads Google's map pack, and it does not describe how it decides which local business to recommend.
What we've seen. Ask ChatGPT "who's the best electrician in Mansfield" and the sources it cites are frequently not company homepages. In the checks we've run for clients we've repeatedly seen directories, review platforms and third-party business listings appearing alongside, and often ahead of, the businesses' own sites.
I want to be straight about the status of that: it's observation from ad-hoc checks, not a structured study, and I'm not going to dress it up as one. It's enough to inform where you spend effort. It is not enough to claim ChatGPT "uses your Business Profile" as a mechanism, and anyone telling you they know the formula is guessing with more confidence than the evidence supports.
Why it's plausible anyway. "Best" is a judgement, and no business's own website is a credible source on whether that business is the best. A tool assembling an answer goes to where other people have said something about you. That's reasoning, not documentation, but it's reasoning that holds up.
So the practical list is unchanged, and it's worth doing regardless of which tool is asking:
- A complete, verified Google Business Profile with the right primary category and the areas you cover. Google separately recommends keeping Business Profile information current for its own AI features, so this one is documented on the Google side at least
- A real, growing pile of genuine Google reviews
- Consistent name, address and phone details across the directories you're listed on, which is the off-page SEO job
None of that is new. It's local SEO, and it was already the thing that mattered most. What's changed is the stakes. Your reputation is no longer just displayed to people who go looking. It's now read, summarised and repeated to someone who never visits any of those pages.
Worth being careful about one thing, though. Google's AI features and ChatGPT's local search are two different systems, and it's tempting to treat them as one because the advice overlaps. Doing well in the map pack is worth chasing on its own merits. Assuming it feeds ChatGPT is a leap nobody has evidence for.
Write pages worth quoting, starting with things only you know
Your own site still matters for the other half of the questions. Not "who's the best" but "how much does a rewire cost", "do I need building regs for this", "how long does a resin driveway take to cure". Those get answered from content, and that content can be yours.
The part that actually differentiates you
If you do one thing from this article beyond the robots check, do this one.
Google's guidance for its AI features comes back to helpful, reliable, people-first content, and warns off content built for machines rather than readers. The practical translation for a trade is first-hand material nobody else can publish:
- Real figures from jobs you've actually completed, not a range copied off another site
- Before and after photos of your own work, with what the problem turned out to be
- The specific complications that come up in your area. Victorian terraces in your town, clay soil, a council that's fussy about dropped kerbs
- Actual timescales, including the bits that go wrong and what that adds
- Why you'd recommend one approach over another, in the words you'd use on someone's doorstep
That's the material a summariser has nothing to substitute for. Every generic "5 signs you need a rewire" article already exists a thousand times over, and a tool assembling an answer has no reason to reach for yours. A page about the job you did in a specific street, with the real cost and the thing you found behind the plasterboard, is far less commoditised. Someone else can write about that street or that kind of job, but they can't have your evidence.
It's also the one recommendation here that holds up whether or not any of this AI stuff pans out. It's just better content.
Formatting, honestly labelled
The usual advice is to answer the question directly under the heading, use headings phrased the way people ask, and put comparative information in tables and lists.
I'd still do all of that, and I've done it in this article. But I want to be honest about why: it makes the page clearer for a human reader. I've seen it presented as a technical extraction mechanism, as though AI tools lift the first sentence under an H2 or prefer tables. I can't find documentation from OpenAI or Google supporting that as a retrieval or ranking factor, and Google's own line is that its systems understand meaning and synonyms without needing exact-match formatting.
So treat it as good writing rather than a trick. Good writing tends to survive contact with summarisers anyway.
One genuine technical point in that list: keep the answer in the HTML. If your content only appears after JavaScript runs, some crawlers won't see it. Worth checking on any site built in a page builder.
On schema markup, use it as ordinary SEO, not as an AI unlock. Google states there's "no special schema.org structured data that you need to add" for its generative features. LocalBusiness markup is still worth having, because it helps Google understand your business details.
Worth knowing if you were told otherwise recently: FAQ markup no longer earns you anything in Google Search. Google retired FAQ rich results in 2026, so the expandable questions that used to show under a listing are gone. Keeping the markup does no harm and it's still valid schema, but nobody should be selling it to you as a Search win any more. And whatever markup you use, don't add anything that doesn't match what's visible on the page.
How to check whether ChatGPT names you
There's no rank tracker for this that I'd trust yet. So do it by hand, and it takes twenty minutes a quarter.
Write down the five questions a customer would genuinely ask before hiring you. Then ask each one, word for word, in ChatGPT. Write down which businesses get named and which sources get cited. Repeat in Gemini, Perplexity and Google's AI Mode if you want the fuller picture.
You'll learn three things fast: whether you appear at all, who your real competition is in this channel (often not who you assumed), and which sources these tools trust for your trade. That last one is the useful bit, because it tells you exactly which listing or review platform to go and fix.
Four things that will ruin your test if you skip them:
- Ask each question several times. Two runs of the same question can produce different businesses, because they were two different searches. You're looking for who shows up consistently, not who appeared once.
- Use a clean session. OpenAI notes that saved memory and personalisation can influence how your question gets rewritten into searches. Your logged-in account knows who you are, which is exactly what a prospective customer's account doesn't.
- Check your location setting. Local answers depend on it, and yours is probably set to where you are rather than where your customers are.
- Write down the date. All of this moves quickly enough that a result without a date is close to useless six months later.
Do it again in three months. That comparison is worth more than any AI visibility dashboard you'll be sold this year.
How to measure it, and the bit nobody mentions
This section changed in June 2026, so ignore anything written before then.
Google: you can now isolate AI impressions
Search Console used to fold AI Overviews and AI Mode into ordinary Web performance with no way to separate them. On 3 June 2026 Google launched Search generative AI performance reports, giving AI visibility its own view for Search and for Discover.
What you get: impressions, plus breakdowns by page, country, device and date. What you don't get, at least at launch: clicks, queries and click-through rate. Like the control, it rolled out to a subset of sites first.
So the honest position is halfway. You can now see whether Google's AI features are surfacing you, which was impossible before. You still can't see what people searched to trigger it, or whether they clicked. Anyone showing you AI-specific query or CTR data from Search Console is showing you something Google doesn't publish yet.
ChatGPT: the tracking is built in
This one is easier than people expect, because OpenAI does the work for you. ChatGPT automatically adds utm_source=chatgpt.com to referral URLs from ChatGPT search, specifically so publishers can identify that traffic in analytics. It's documented in OpenAI's publishers and developers FAQ.
So in GA4, filter or build an exploration on Session source equal to chatgpt.com. That's cleaner than eyeballing the referral list. GA4 also added an AI Assistant default channel group that recognises sources like ChatGPT, Gemini and Claude and buckets them for you, so you can often just read that channel straight off the acquisition report.
Set that up once and you have a running count of how often ChatGPT sent someone your way.
What this actually looks like, on our own site
Rather than tell you the numbers are small, here are ours.
In the 90 days to 14 August 2026, watwebsites.co.uk had 1,736 sessions. Of those, 45 came from chatgpt.com, about 2.6%. Every single session GA4 filed under its AI Assistant channel came from ChatGPT. Nothing from Perplexity, Gemini, Copilot or Claude registered at all.
Two things to take from that, and one caution.
It's small. Anyone telling a plumber to reorganise their marketing around AI search this year is ahead of the evidence.
It's also not nothing, and it's concentrated. For this site, in this period, ChatGPT was the AI channel. If you only have the appetite to check one tool, that's a reasonable steer on which.
The caution: this is our own site, and we sell websites and SEO. Our visitors are far more likely to be ChatGPT users than someone hunting for an emergency plumber at 9pm. Don't read our 2.6% as your 2.6%. Go and look at your own AI Assistant channel, which takes about a minute.
The gap nobody sells you a fix for
When ChatGPT names your business and the customer rings you straight off the back of it, nothing lands in your analytics at all. The recommendation worked and left no trace. No referral, no UTM, no session. Worth knowing before you judge this channel purely on the numbers above, and worth asking new enquiries where they heard about you.
We wrote about the wider effect of this shift, with real Search Console data from one of our clients, in Google's AI Overviews are eating clicks. Short version: impressions up, clicks down, and the businesses that only ever counted clicks are the ones panicking.
The honest summary
If you run a service business in the UK, here's the whole thing.
For local and current questions ChatGPT can search the web before it writes, so being findable is the strategy for when it does. Check nothing is blocking OAI-SearchBot, and check your firewall as well as your robots.txt, because that's the one job genuinely unique to this. Google says plainly there are no special AI optimisations, no magic files and no special schema, so treat anyone selling those accordingly. Then put your effort into your Google Business Profile, your reviews, your listings, and content built on jobs you've actually done.
That's it. It's mostly the SEO you should already be doing, done properly, plus one file check.
The reason it's worth doing now isn't traffic. On our own site ChatGPT accounted for 2.6% of sessions over the last 90 days, and we're an audience unusually likely to use it. Yours will probably be lower. It's worth doing because the businesses that get quoted in these answers are being recommended rather than just listed, and that's a materially different thing to be. It's also worth doing because almost all of it is work that pays off in ordinary search regardless of whether any of this becomes a real channel.
Frequently Asked Questions
How do I get my website to show up in ChatGPT?
Start by checking nothing blocks OAI-SearchBot, the crawler OpenAI uses to surface sites inside ChatGPT's search features. That means your robots.txt and also your firewall, CDN or bot protection, which can refuse a crawler before robots.txt is ever read. After that it's ordinary SEO, because for local and current questions ChatGPT searches before it writes: content that answers the question clearly, a complete Google Business Profile, and genuine reviews and listings that confirm you exist. There's no ChatGPT ranking to game, because there's no fixed ranking.
Does ChatGPT use my website or its training data to answer?
For local and current questions it generally searches. OpenAI says ChatGPT searches the web when a question might benefit from current information, retrieving live pages and profiles and then summarising and citing them. That's why being findable matters more than any phrasing trick. Search visibility and model training are controlled separately and serve different purposes, so being in training data isn't what gets you named in an answer.
Do I need an llms.txt file to appear in ChatGPT or AI search?
No. Google updated its documentation in June 2026 to say that llms.txt files aren't used by Search, including its generative AI features, and that they neither help nor hurt rankings. If someone is charging you to add one as an AI optimisation service, that's the moment to ask what else they're selling.
Does blocking AI crawlers hurt my Google rankings?
It depends which crawler. Blocking Googlebot stops Google crawling the page, which effectively destroys its useful Search and AI visibility, though a blocked URL can sometimes stay indexed if Google finds it through links. Blocking Google-Extended only limits how your content is used for Gemini and other generative products, and doesn't affect your eligibility in Search or its AI features. Blocking OpenAI's or Perplexity's crawlers has no effect on Google at all, it just removes you from those tools.
Can I see how much traffic I get from ChatGPT?
Yes, on the ChatGPT side. OpenAI automatically adds utm_source=chatgpt.com to referral URLs from ChatGPT search so publishers can identify that traffic, so in GA4 you can filter Session source for chatgpt.com, and GA4's AI Assistant channel group buckets it for you. On the Google side it's partial: Search Console added dedicated Search generative AI performance reports in June 2026 which isolate AI impressions by page, country, device and date, but they don't yet include clicks, queries or click-through rate. Neither captures the customer who sees your name in an answer and simply rings you.
Is AI SEO or GEO a different discipline from normal SEO?
Mostly not. Google states there are no additional requirements to appear in AI Overviews or AI Mode and no special optimisations necessary. The genuinely new parts are narrow: deciding which AI crawlers you allow, and writing pages that answer a question directly enough to be quoted. The rest of what gets sold as generative engine optimisation is the SEO you should already be doing, with a new invoice attached.
Should a local service business worry about ChatGPT yet?
It's worth an hour, not a budget. UK search volume for AI visibility terms is still small and most local enquiries still come through normal search and the map pack. The sensible position is to spend an hour making sure you're not accidentally blocked or invisible, keep your Google Business Profile and reviews strong because that's what these tools quote for local questions, and revisit it each quarter.
Want to know whether ChatGPT is naming you or your competitor? We'll run the questions your customers actually ask, show you who's getting cited and why, and go through what's worth fixing on a quick call. Book a free call here.

Written by
Will Gadsby, Co-founder, WAT Websites
Will co-founds WAT Websites, where the team builds fast, custom websites and runs local SEO for UK service businesses, so their phone rings more often.