How to get your business recommended by ChatGPT

Open ChatGPT tonight and type the question your best customer would type. "Who's the best bakery near Indiranagar?" "Which accountant in Pune handles small exporters?" Whatever it is for you.

Read the answer slowly. Two or three names come back, each with a tidy little reason. If one of them is yours, lovely. If none of them is, you've just met the newest gatekeeper in your sales process, and it never asked for a meeting.

ChatGPT and Google's AI Mode recommend businesses by searching the web and summarising what they find: your website, your listings, your reviews and what others write about you. You can't buy that answer, but you can make sure every source it reads is accurate, consistent and easy to find. This guide gives you the test, the checks and the fixes, in the order that matters.

Featured image: AI-generated illustration of a fictional bakery owner, not a client. Business names and prompts in this article are made-up examples. I build and fix websites for a living, so I have an interest here; every claim links to its source so you can check me.

Do customers really ask ChatGPT for business recommendations?

Yes, and the jump happened fast. In BrightLocal's 2026 survey of 1,002 US adults, 45% said they now use AI for local business recommendations, up from 6% a year earlier. ChatGPT was the most used tool at 31%, with Google's AI Mode at 23%.

I spent twenty years selling capital market products before I wrote code, and the sale rarely happened on the brochure. It happened when the client rang someone they trusted and asked, "Who would you use?" That phone call has moved into a chat box. The chat box doesn't owe anyone a favour.

The numbers worth knowing, and what each one measures

Most AI search statistics come from companies that sell AI search tools. These are the ones I could trace to the original source, with the catch attached.

Read together, they say something simple. Customers are asking. The shortlist is tiny. And they still check the reviews before they pick up the phone.

How does ChatGPT decide which businesses to recommend?

It searches, reads and summarises. When a question needs fresh information, the assistant turns it into web searches, reads the pages that come back, and writes an answer from them. OpenAI runs a separate crawler for this, OAI-SearchBot, which "is for search" and is what surfaces websites in ChatGPT's answers.

The AI doesn't know your business. It knows what the web says about your business, which is not always the same thing.

Nobody outside OpenAI or Google knows the exact ranking rules, and anyone who claims to is selling something. What we do know is where the raw material comes from. BrightLocal notes that AI answers draw on reviews, local directories, business websites, social media and other third-party sources, depending on the tool and the question.

Is optimising for AI different from normal SEO?

Mostly no, according to Google itself. Its guide to optimising for generative AI features, published in May 2026, says "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." Its AI features page adds that there are "no additional requirements to appear in AI Overviews or AI Mode."

Here's where I disagree with half the internet. "GEO" and "AEO" (generative and answer engine optimisation) are mostly good SEO plus a good reputation, with new names that justify new invoices. The parts that genuinely matter are older and duller than the pitch decks suggest.

Step 1: find out what AI says about you today

Run five questions through three assistants and write down what comes back. It takes about twenty minutes, costs nothing, and gives you a baseline so you can tell later whether anything you changed worked.

Use a fresh chat with no history, ideally in a private browser window, because assistants can tailor answers to what they already know about you. Run each question twice. Answers vary between runs, so one bad answer is a data point, not a verdict.

Test in ChatGPT, Google (AI Mode or the AI Overview at the top of results) and one more, such as Gemini or Perplexity. Then fill in one row per answer:

The "sources it cited" column is the gold. If every answer cites the same directory, review site or local blog, that's where your effort should go first, not your homepage.

Step 2: make sure AI search can read your website

Check that your website isn't quietly locking the door. A single line in a file called robots.txt can hide your whole site from ChatGPT's search, and it's often set by a plugin or a website builder setting that nobody remembers switching on.

Type your domain followed by /robots.txt into a browser, for example www.yourbusiness.com/robots.txt. You're looking for any rule that blocks OAI-SearchBot, Googlebot or every crawler at once (User-agent: * followed by Disallow: /).

Blocking AI training is fine; blocking AI search is not

OpenAI treats these as separate switches. Blocking GPTBot keeps your pages out of model training, while OAI-SearchBot controls whether you appear in ChatGPT's search answers, and OpenAI says each setting is independent. Sites that opt out of OAI-SearchBot won't appear in ChatGPT search answers.

So you can say no to training and yes to being recommended. This is a working robots.txt that does exactly that:

# Let ChatGPT search find and cite the site
User-agent: OAI-SearchBot
Allow: /
 
# Keep pages out of OpenAI model training (does not affect search)
User-agent: GPTBot
Disallow: /
 
# Everyone else, including Googlebot and Bingbot
User-agent: *
Allow: /
 
Sitemap: https://www.example.com/sitemap.xml

I ran this file through Python's standard robots.txt parser: OAI-SearchBot, Googlebot, Bingbot and PerplexityBot were allowed to fetch a services page, and GPTBot was refused. OpenAI says changes take about 24 hours to register. Replace the sitemap address with your own, and don't paste this over an existing file without checking what the old rules were protecting (admin pages, carts and the like).

Google's side: indexed, with a snippet

For Google's AI Overviews and AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet. A noindex tag removes the page from Search entirely, and nosnippet limits what Google can show from it. Both are easy to set by accident and hard to spot by eye.

The message to send your web developer

If someone else looks after your site, send them this. It asks for evidence, not reassurance:

Please check three things on our website and reply with screenshots:

1. Our robots.txt allows OAI-SearchBot, Googlebot and Bingbot to read our homepage and service pages. Blocking GPTBot is fine.
2. Google Search Console's URL Inspection shows our homepage and main service pages as indexed.
3. None of those pages use noindex or nosnippet.

Also tell me whether our website builder or any plugin has a setting that blocks AI crawlers, and what it's set to.

A good developer answers this in under an hour. If the reply is a paragraph of reassurance with no screenshots, ask again.

Step 3: fix the facts AI reads about you elsewhere

Your website is one source among many, and often not the one the AI quotes. Listings and reviews do a lot of the talking, which is good news, because they're the cheapest part to fix.

Claim and complete your Google Business Profile

Google's own guide names Business Profiles as a way to help your products and services be visible in AI responses. Fill in every field that's true: categories, service area, hours (including holidays), services with plain descriptions, photos of the real place. An empty profile is like a shop with the shutters half down. People walk past.

Make your name, address and phone match everywhere

SOCi's study of brand locations found that "inconsistent or incomplete listings reduce AI confidence". If one directory has your old address and another lists a number you gave up in 2022, the AI has to guess which one is true. It may guess the other bakery.

List the five directories your Step 1 answers cited most, plus the obvious ones for your industry. Correct each one so the business name, address, phone, hours and website are word-for-word identical. Boring work. It pays.

Earn reviews steadily, and never fake them

In SOCi's data, locations ChatGPT recommended averaged 4.3 stars. That study covered chains, so treat the exact number with care, but the direction matches what customers say: BrightLocal found 97% of AI users sometimes double-check AI recommendations against real reviews.

Ask every customer for an honest review, with a direct link, once the job is done. Every customer, not just the ones who smiled: Google's review policy says businesses must not "selectively solicit positive reviews" or offer incentives for reviews, and it can suspend accounts that do. Don't pressure people to review while they're still in the shop, either. Reply to reviews, especially the bad ones, calmly and specifically.

Don't buy reviews and don't write them. In the US, the FTC's rule bans fake reviews, including AI-generated ones, and rewards conditioned on a positive review, with civil penalties for knowing violations. Other countries have their own consumer rules, so check yours before you hire anyone to "manage" reviews.

Step 4: write the page that answers the buyer's question

Give each main service its own page that answers what a buyer asks before they buy. Google's guide rewards content you create from what you know, with "unique expert or experienced takes that go beyond common knowledge". A page that could belong to any business in your category gives an AI nothing to quote.

The examples are illustrative, for a fictional bakery. Use your real numbers, and only claims you can stand behind.

If you can't write the specific version in one sentence, the problem is upstream of the website. The branding guide for people who hate branding has a quick exercise for finding that sentence.

You don't need to rewrite your site in some special robot dialect. Google says plainly that there's "no requirement to break your content into tiny pieces" and you "don't need to write in a specific way just for generative AI search". Write for the customer. The AI reads over their shoulder.

What doesn't work (and what people will try to sell you)

Most paid "AI visibility" shortcuts target things Google says it ignores. Before you sign anything, hold the pitch against the primary source:

A small confession: this site has an llms.txt file. It costs nothing to keep, and other tools may read it. But it isn't why anything here gets found, and I'd never charge a client for one. Structured data still earns its place for a different reason: Google says it helps pages qualify for rich results in normal search.

How do you know if it's working?

Measure enquiries, not mentions. AI answers often mean fewer clicks: in Pew's data, people clicked a link inside the AI summary on just 1% of visits. So website visits are a weak scorecard. Visits can fall without enquiries falling with them, and enquiries are the number that pays rent.

Three signals are worth tracking:

• Your prompt test, monthly. Rerun the Step 1 sheet on the same day each month. Count how often you're named and how often the facts are right.
• Google Search Console. The Search Generative AI performance report, rolled out to all websites by 31 August 2026, shows how often your pages appeared in AI Overviews and AI Mode, page by page. It counts impressions. Clicks from those features stay blended into the main Performance report under "Web", so that trend can't tell you on its own whether AI visibility improved.
• The question every new customer gets. Add "ChatGPT or another AI assistant" to your "How did you hear about us?" options. Software misses a lot; people remember. The guide to traffic that turns into pipeline explains how to tie answers like these to actual revenue.

The work is a loop, not a launch. Each monthly retest tells you which step to go back to:

Try this: one evening, four checks

Block out an evening this week and do these in order. Each one ends with something written down, so you'll know exactly where you stand.

1. Run the prompt test from Step 1 across three assistants and save the sheet.
2. Open your robots.txt and confirm OAI-SearchBot and Googlebot aren't blocked, or send the developer message from Step 2.
3. Open your Google Business Profile and fix one wrong or empty field. Then fix the most-cited directory from your sheet.
4. Read your main service page and rewrite one vague sentence into a specific one, using the table in Step 4.

It worked if, a month from now, your rerun of the same five questions shows fewer wrong facts and at least one new source citing you correctly. Being named takes longer. Accurate comes first.

Frequently asked questions

Can an agency guarantee my business shows up in ChatGPT?

No. Nobody outside the AI companies controls which businesses an assistant names, and even Google says indexing and serving isn't guaranteed for pages that qualify. An honest agency promises the work (crawlable site, accurate listings, a review process, a useful service page) and reports what changed, not a ranking.

Do I need an llms.txt file for AI search?

Not for Google. Google's guide says its Search ignores llms.txt and the file will neither harm nor help your visibility. It's harmless to keep if you have one, but don't pay anyone to make it.

Should I block AI crawlers to protect my content?

Block training if you like, but not search. OpenAI keeps the two separate: disallowing GPTBot opts you out of training without affecting ChatGPT search, while blocking OAI-SearchBot removes you from its search answers. If you want to be recommended, OAI-SearchBot has to be allowed in.

How long does it take for ChatGPT to start recommending my business?

There's no fixed timeline. OpenAI says robots.txt changes take about 24 hours to register, but listings, reviews and other sites mentioning you build up over months. Expect accurate facts first and recommendations later, and judge progress with a monthly rerun of the same questions.

Is GEO or AEO different from SEO?

Very little, by Google's account. Its own guide says optimising for generative AI search is "still SEO". The difference in practice is emphasis: reviews, listings and third-party mentions matter as much as your own pages, because the AI reads all of them.

Do these numbers apply outside the US?

The surveys here come from the US, North America, the UK and Australia, so treat the percentages as a sign of direction, not a local fact. The mechanics are the same everywhere: the assistant searches, reads your site, listings and reviews, and summarises. Run the Step 1 test in your own city to see what's true for you.

Be the business the evidence points to

AI assistants don't pick favourites. They repeat whatever the web says most clearly and most consistently. Your job is to make sure that's the truth: a site they can read, listings that agree with each other, reviews from real customers and a service page with real answers. Start with the twenty-minute test tonight, because you can't fix an answer you haven't read.

If you'd like a second pair of eyes on your results, send me your prompt test sheet and I'll tell you which fix I'd make first, even if it's one you can do yourself. You can also see how I approach digital marketing work.