How AI search chooses which businesses to recommend

One of the most common questions we're being asked now is: “how do I get my business recommended by ChatGPT?”

It's a sensible question. People are starting to use AI assistants to do things they would previously have done through Google, like asking for a good independent BMW specialist near Eastbourne, which company should build an automotive ecommerce website, or where to buy performance parts for a Golf GTI.

And instead of receiving ten blue links, they may receive a shortlist, summary or direct recommendation. So why does one business appear while another doesn't?

There is no single answer. AI search systems can use a combination of web pages, business information, reviews, search indexes, third-party sources and other signals to construct an answer. But there are some clear principles businesses can work on.

The short version

AI systems are more likely to surface a business when they can clearly understand what it does, where it operates, who it serves and what evidence exists to support its expertise. There is no guaranteed way to “rank number one in ChatGPT”. The practical job is to make your business easier to discover, understand and verify across the web.

This is an important misconception. AI assistants do not necessarily work like a directory containing a fixed list of companies ranked from best to worst.

The answer can depend on the exact question, the user's location, the type of business, the information available online, which sources are retrieved, how current those sources are, and the platform being used.

OpenAI says ChatGPT Search ranks results using multiple factors intended to surface relevant and reliable information, and that placement is not guaranteed.

That means the real objective isn't “how do I become ChatGPT's favourite business?” It's “how do I make my business a strong, credible candidate when somebody asks a relevant question?”

You're not trying to convince an AI that you are the best business in the world. You're trying to give it enough reliable information to understand when you are a relevant answer.

Imagine somebody asks: “who is a good garage near Brighton for servicing a modified Golf GTI?” There may be dozens of garages nearby, but the most relevant candidate isn't necessarily the closest.

A website that clearly discusses Volkswagen servicing, Golf GTI models, modified vehicles, performance upgrades, diagnostic capability and track cars provides much stronger context than a website saying “we service all makes and models.”

Google uses a similar principle in local search. Its own guidance says local results are primarily influenced by relevance, distance and prominence, and that complete, detailed business information helps Google understand when a business matches a search. AI search makes that need for context even more obvious.

If these four things are vague, contradictory or missing, it becomes much harder for any search system to understand the business.

This is one of the biggest issues we see on websites. A homepage says “helping ambitious businesses achieve their goals through innovative solutions.” It sounds polished. But it tells a search engine almost nothing.

Compare it with “we design and market websites for independent garages, automotive businesses and performance parts companies across the UK.” Now we know what the company does, who it works with, which industries it understands and where it operates. That clarity is valuable for customers as well as machines.

This is probably the most important part of the article. Search platforms can build an understanding of a business using information from across the web.

Google explicitly says its Business Profiles can use information from official business websites, publicly available web content, licensed third-party data, business owners, user contributions, reviews and photographs.

That means your online presence is bigger than your website. An AI system may encounter your website, Google Business Profile, review platforms, news articles, directories, social profiles, partner websites, industry associations, client websites, product listings and forum discussions. This is why consistency and reputation matter.

One story, everywhere

Your website cannot say one thing while the rest of the web says another. Imagine your website says Eastbourne, your Google Business Profile says Polegate, an old directory says Hailsham, and Facebook hasn't been updated for five years.

A human might work it out. A search system has to reconcile conflicting information. The clearer and more consistent your presence is, the easier that job becomes.

Anyone can write “we are automotive website experts.” That doesn't prove anything. But now imagine the website also contains automotive case studies, photographs from garages and workshops, ecommerce projects, client testimonials, specialist automotive articles, named clients and real project results. Suddenly the claim has context.

This is where AI visibility and conventional marketing overlap nicely. The things that help a potential customer believe you are credible are often the same things that help search systems build a stronger understanding of you.

For local businesses, reviews can be particularly important. Google says review count and review score can contribute to local prominence, alongside information such as links and references across the web.

But I wouldn't think only in terms of “we need more five-star reviews.” The content inside reviews can also provide useful context.

The second tells us considerably more about the vehicle, the service, the capability and the customer experience. The same is true for human readers.

This is one of the biggest opportunities for smaller businesses. You don't need to be the biggest company. You need to be highly relevant.

Suppose somebody asks: “who builds ecommerce websites for motorsport parts companies?” A huge general agency might have 100 staff, hundreds of clients and a very strong domain. But if its website contains almost no evidence that it understands automotive ecommerce, another specialist business may be a much better match.

That's why niche expertise can be so valuable in AI search. And it's why content helps. Not publishing articles purely to feed AI, but answering the questions that naturally surround your work. For a garage, that could be unusual faults, specialist servicing, workshop projects, technical comparisons and common mistakes.

Being smaller doesn't stop you becoming the most relevant answer to a very specific question.

Example: two garages

Both might be excellent garages. But Garage B gives a search system far more information with which to understand its expertise.

None of this works if search platforms cannot access the website. OpenAI recommends allowing OAI-SearchBot if a publisher wants its content to be discoverable and surfaced within ChatGPT Search. Likewise, conventional search engines still need to crawl and index important pages.

So some very traditional technical checks remain important: robots.txt, indexability, canonical URLs, internal links, XML sitemaps, page rendering, mobile usability and site structure. There's no point publishing brilliant evidence if it's buried somewhere search systems cannot access. We cover the groundwork on our SEO & search visibility page.

Potentially. Structured data can help machines identify things such as organisations, local businesses, people, products, reviews and articles. But it shouldn't be treated as a shortcut. The markup needs to reflect information that's really on the page.

Schema can help describe the evidence. It cannot manufacture the evidence.

A useful way of thinking about this: what would somebody else find if they researched your business? Do credible third parties mention you? Think industry bodies, trade associations, manufacturers, partners, customers, suppliers, news publications or specialist directories?

Google specifically says prominence can be influenced by what it knows about a business from across the web, including links and references. You don't need hundreds of meaningless directory listings. A small number of genuine, relevant references can tell a much stronger story.

Social media is worth a brief word too. I wouldn't claim that posting every day on Facebook suddenly makes ChatGPT recommend a business. There's no evidence for that. But social profiles can contribute to the wider picture: that the business is active, that real people work there, recent work, locations and customer interaction. And customers themselves may encounter those profiles while researching you. The objective isn't “post on Instagram for GEO”; it's building a credible, consistent digital presence.

Recommendation is only one type of AI visibility. Your website may also be used as a source. Microsoft now provides publishers with AI Performance reporting in Bing Webmaster Tools showing how often pages are cited, which URLs are cited, and the queries associated with those citations.

That's important because it shows how measurement is starting to evolve. We're moving from only asking “where do we rank?” towards also asking “where is our information being used?”

Think like the person asking the question

If somebody asks “who should I use?”, what information would they need before they felt comfortable recommending you? That's a very good starting point for AI search optimisation too.

No reputable agency should promise “we'll make ChatGPT recommend you.” OpenAI itself states that placement in ChatGPT Search is not guaranteed.

Different AI systems behave differently. Their search technology changes. The sources they retrieve can change. And the question asked changes the answer.

The useful work is therefore about improving eligibility and relevance, not claiming control over the final response.

Be wary of guaranteed “ChatGPT rankings”

There is a real opportunity around AI search. There is also a lot of opportunistic marketing.

If someone promises they can guarantee your company will become the number-one recommendation in ChatGPT, ask them exactly how that guarantee works. There is currently no equivalent of buying the number-one organic position.

For most businesses, it looks surprisingly similar to good digital marketing: a technically sound website, clear services, accurate business information, specific expertise, genuine reviews, useful content, real-world evidence, credible third-party references and good measurement.

That doesn't sound quite as exciting as a secret GEO hack. But it is much more useful. If you'd like the fuller picture of how the terminology fits together, we covered that in SEO, AEO and GEO: what's the difference? and in is your website ready for AI search?

OpenAI — Searching the web with ChatGPT. Explains that ChatGPT Search uses multiple factors to identify relevant and reliable results, and that placement is not guaranteed.

OpenAI — Publishers and Developers FAQ. Guidance for website owners covering OAI-SearchBot, website discovery and referral tracking from ChatGPT Search.

Google Business Profile — Tips to improve your local ranking on Google. How relevance, distance and prominence contribute to local search visibility.

Google Business Profile — How Google sources business information. How website content, Business Profiles, third-party data and user contributions are combined.

Microsoft Bing — AI Performance in Bing Webmaster Tools. Reporting that shows how website content is cited within Copilot and Bing's AI-generated experiences.