Why Doesn’t ChatGPT Recommend My Business? (And How to Fix It)

The question every owner started asking in 2026

A dentist in Lisbon, a SaaS founder in Berlin, a lawyer in Manchester. Different businesses, same email to us this year: “I asked ChatGPT for the best [my thing] in [my city] and it listed three competitors. Why not me?”

The reason it stings is that the question behind it changed. For twenty years the buyer journey started with a Google search and ten blue links, and your job was to be one of the links. Now a growing share of buyers ask an assistant and get one answer. Not a list to browse. A recommendation. If you are in it, you exist. If you are not, the buyer never learns you were an option.

We have been measuring this in our own market since early 2026. Every month we run the questions real buyers ask about AI coding tools across ChatGPT, Claude and Perplexity and publish the results as a public recommendation index. The pattern is consistent: the engines converge on a short list of names, the same names, month after month. Everyone else splits the leftovers. The mechanics that decide that short list are the subject of this article.

How AI assistants actually decide who to recommend

There is no secret ranking algorithm to reverse-engineer. AI assistants answer from two inputs: what they learned in training, and what they retrieve live from the web when they answer. Both inputs are made of sources. Comparison articles. Industry directories. Review platforms. Reddit threads. Documentation. Pages that state facts plainly enough to be quoted.

When Perplexity answers “best CRM for a small law firm,” it cites the pages it built the answer from, and you can read the list. ChatGPT and Claude are less transparent about it, but they drink from the same well. In our monthly index we track exactly which domains the answers lean on in our category, and it is a short, stable list. Every market has one. We call it the cited-source list, and whether you are on it is the single biggest factor in whether you get recommended.

Here is the uncomfortable part, straight from our own published data: being the best product does not put you in the answers. Our index compares recommendation share against public benchmark scores, and the sharpest divergence we found was a tool statistically tied for the top score on its category benchmark that received zero recommendations in the developer answers that month. Best on merit, invisible in the conversation. Meanwhile the most-recommended tool in the category shows up in roughly one of every three answers.

If that happens to well-funded software companies in the most-watched category on the internet, it is happening to your business too. The engines do not know how good you are. They know what the sources say.

The four reasons your business is invisible to AI

After auditing sites across coaching, finance, local services and SaaS, the same four problems come up in nearly every invisible business. Most have all four at once.

What actually moves the needle

The fixes mirror the causes, and none of them are secret. They are just work, in priority order.

First, find and enter the cited sources for your market. Ask Perplexity the ten questions your buyers ask and read the citations. That list of domains is your target list. Some you can join today: directories that take submissions, review platforms where you can claim a profile and earn reviews, communities where your customers already talk. Some require outreach: the roundup articles and industry publications the engines keep quoting. Getting into three of the right sources beats a hundred backlinks from sites the engines never read.

Second, publish answerable pages. One page per real buyer question, with the question in the title, a direct answer up top, the reasoning below, and a visible date. Pricing pages that state prices. Comparison pages that name your competitors honestly, because the engines cross-check and a one-sided page reads as marketing. This is exactly the structure we use on our own site, and it is why our pages show up as citations in our own category’s answers.

Third, ship structured data. Organization, LocalBusiness or Service, FAQPage, Product, Review: whichever types match what you sell, implemented in JSON-LD, validated, on every page that matters. It is a few days of work and it is the cheapest visibility gain on this list.

Fourth, measure it like a channel. You would not run ads without conversion numbers. Run the same buyer questions across the engines monthly, log who gets named, and watch your share move as the work lands. That feedback loop is what turns this from superstition into marketing.