How to get your business mentioned in ChatGPT's answers

by Harry Watts · 24 September 2026 · 8 min read

ChatGPT Ads can put your business in a labelled slot beside an answer, but not in the answer itself. A mention there has to be earned: ChatGPT has to find you, read you and trust you enough to name you. That comes down to four things you control: letting its search crawler in, being in the search indexes it draws on, publishing specific facts worth quoting, and being described the same way everywhere you appear.

None of that is new. It's search engine optimisation, pointed at a different reader.

Google went this way first

I'd explain this to any business owner with Google. Organic results came first, then came paid ads: Google launched in September 1998, AdWords in October 2000. ChatGPT has followed suit. It launched in November 2022, and its ads went live in the UK in June 2026.

So the split is one you already know from a Google results page. The ChatGPT Ads slot is a labelled card, bought at auction, and OpenAI's policy bars ads from influencing what the assistant says. The answer itself is the organic result: nobody pays for it, and it goes to whoever the system judges most useful for the question. I manage the paid half for clients, and I'm running the £500 credit offer on Search Lucid's own account. None of that buys a word of the answer.

So the question "how do I get into ChatGPT?" has two answers, and it's worth knowing which one you're asking about before you spend anything.

How ChatGPT decides who to mention

An assistant can know about your business in two different ways, and they behave nothing alike:

The right-hand column is where a business can act this month. Questions about current prices, local businesses and recent comparisons ("best accountant in Leeds", "how much does a loft conversion cost") are the ones training data can't answer reliably, and a search is something you can be found in.

Step one: let the search crawlers in

Each AI company runs separate crawlers for separate jobs, and robots.txt treats each one independently. The ones that fetch pages to answer a question and cite the source are the ones that decide whether you can appear at all:

OpenAI's own documentation is blunt about the first one: sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. Block it and nothing else in this article matters.

Most sites never block it on purpose. They block it by accident, in one of three ways:

• A blanket AI block. A plugin or a template adds Disallow rules for "AI bots" and catches the search crawlers along with the training ones.
• A CDN setting nobody chose. Cloudflare has blocked AI crawlers by default on new domains since July 2025. A firewall block never appears in your robots.txt, so the file can say "allowed" while the crawler is turned away.
• An edited file that isn't the live one. CDNs and hosts can add their own rules to robots.txt on the way out.

So check what's actually served, not what's in your CMS. Open yoursite.co.uk/robots.txt in a browser, and if you use Cloudflare, look at its AI crawler settings in the dashboard.

Blocking the training crawlers is a separate decision

GPTBot, ClaudeBot and Common Crawl's CCBot collect text that may be used to train future models. OpenAI states that its settings are independent: you can allow OAI-SearchBot to appear in search results while disallowing GPTBot. Anthropic separates ClaudeBot from its search and user crawlers in the same way.

Whether to block them has real arguments on both sides:

Google is the awkward case. Google-Extended doesn't affect Google Search or AI Overviews, but Google's documentation says it also covers grounding in Gemini's apps, meaning whether Gemini can use your pages at the moment it answers. Blocking it is not free.

So should you block them? It's a very nuanced topic, and I'd still tell a client to do their own research and decide for themselves. The table above is where that research starts. But I'll tell you which way I went: on this site, I let them in. The content here isn't the product, and I need the brand exposure, so in my opinion it's worth the risk. If your writing is what you sell, the same reasoning points the other way. No AI company publishes how much training data shapes which businesses get recommended, so there's no figure to settle it either way.

Step two: be in the indexes ChatGPT searches

ChatGPT doesn't search the web from scratch every time. OpenAI says its search uses third-party search providers, and Bing is the one it names. If Bing hasn't indexed a page, that route to ChatGPT is closed.

• Bing Webmaster Tools. Verify the site and submit your sitemap. It takes minutes, and most small businesses have never done it because Bing's share of UK searches (about 6%) never seemed worth the trouble.
• IndexNow. A free protocol that tells Bing about new and changed pages straight away instead of waiting for a crawl.
• Google Search Console. Google's index feeds AI Overviews and Gemini, so the same basics apply there.
• Pages that work without JavaScript. Many AI crawlers read the page source without running scripts. Content that only appears after JavaScript runs may never be seen.

Step three: publish things worth quoting

An assistant building an answer is looking for a sentence it can lift and attribute. Vague pages don't give it one. What gets quoted:

• The answer in the first paragraph. Not after an introduction about why the question matters.
• Specific, checkable facts. A price range, a timescale, a number with a unit. "Prices vary" is useless to an assistant; "£200–£800 a month" is a sentence it can cite. The Google Ads cost article is built this way.
• Tables written in HTML. A comparison in a table is easy to extract. The same comparison in an image is invisible.
• The exact names of things. Products, settings, places, standards. Assistants match questions to named entities, so "Consent Mode v2" beats "the new cookie rules".
• A visible date. A reader, and an assistant choosing between two sources, has more reason to trust a page that says when it was last checked.

Step four: be the same business everywhere

Before naming a business, an assistant has to be confident it's one real, specific business. It builds that confidence from agreement: the same name, the same description and the same facts on your site, your Google Business Profile, LinkedIn, directories and review sites.

• Structured data. Organization or LocalBusiness markup on your site, with sameAs links to every profile that's genuinely yours. That's the machine-readable version of "these are all the same business".
• One description. If your site says "boutique accountancy" and your directory listing says "bookkeeping services", you've given the assistant two businesses to reconcile.
• Mentions you don't write. Reviews, press coverage, a directory listing, a useful answer in a forum. Your own site is only your word; independent mentions are evidence an assistant can check. They're also why this step is the slowest.

What about llms.txt?

llms.txt is a proposed standard: a plain-text file at the root of your site that summarises it for AI systems. This site has one. But no major AI company has said it reads the file, and Google's John Mueller has said none of them do. It's cheap to publish and does no harm, but it isn't what gets a business cited.

How to tell whether it's working

No dashboard reports how often an assistant names you. Two things get close:

• Referrals in GA4. ChatGPT adds utm_source=chatgpt.com to many of the links it cites. In GA4, go to Reports → Acquisition → Traffic acquisition and search for "chatgpt". Some clicks arrive without it and show up as direct traffic instead, so treat the figure as a minimum.
• A fixed set of questions, asked monthly. Write down ten to twenty questions your customers actually ask. Once a month, ask each one in ChatGPT, Claude, Perplexity and Google's AI Mode, in a fresh conversation, and log whether you were cited with a link, mentioned without one, or absent. Same questions every month, or the trend means nothing.

Expect it to be slow. Search indexing takes days to weeks, and the reputation step takes months. Anyone promising a guaranteed placement in ChatGPT's answers is promising something nobody outside OpenAI controls.

Quick answers

If you'd rather be seen in ChatGPT while the organic side builds

Everything above takes months to pay off. The paid half works the week it launches, if your business is one ChatGPT Ads can serve. Thirty minutes on a free intro call settles whether it is, and whether the £500 credit is worth taking for you.