Why Does ChatGPT Recommend My Competitors Instead of Me?

It’s very easy to look at that list and think they must have done something you haven’t. Maybe their SEO is stronger, their website is better optimised for AI, or they’ve found some trick you missed. And once you start comparing their sites with yours, there is no shortage of differences to find.

The problem is that the recommendation doesn’t tell you which of those differences, if any, had anything to do with the result.

When the same competitor appears repeatedly, you have something concrete to investigate. You still don’t have the explanation.

OpenAI doesn’t publish a neat business-recommendation formula that tells us exactly why Business A appeared and Business B didn’t.

For ChatGPT Search, OpenAI says ranking is based on a number of factors intended to help people find reliable, relevant information, and that there is no way to guarantee top placement.[1] That’s about as close as we get to an official high-level explanation.

So if one competitor appears once, I wouldn’t immediately start reverse-engineering their website.

If the same competitor keeps appearing across several closely related questions, though, the pattern becomes much more interesting. Now there is something to compare because ChatGPT keeps connecting that business with the need being described.

Even then, you can see what ChatGPT keeps doing, but you still don’t know why.

And this is where competitor analysis can go off the rails remarkably quickly. You notice that the other business has more blog posts, 300 reviews, a different page title, 4 directory listings you don’t have, schema you haven’t installed and a founder who has been in business since the dawn of broadband. Before long, you have a shopping list of changes and still no evidence that any one of them explains the recommendation.

A better comparison starts much closer to the question ChatGPT was actually asked.

In another article, I looked at the difference between asking ChatGPT about a business by name and asking it to recommend a business for a particular need. Competitor comparison only makes sense in that second kind of search.

Suppose someone asks:

Once the query is clear, I would compare the way both businesses describe the relevant work.

I’m not looking for prettier copy. I’m looking at whether somebody unfamiliar with either business could work out, quite quickly, what each one actually does and where it fits.

I'd look at:

This is one of those places where ordinary marketing language can get in the way. A business owner knows perfectly well that “strategic transformation for sustainable growth” includes succession planning. A search system, a prospective client, and frankly quite a few humans have to work harder to make that connection.

If the competitor consistently uses the language of the service people are asking for, and your business describes the same work in much broader terms, I would want to check whether that broad wording is making your service harder to connect with the search.

We still don’t have a ranking formula, but we do have evidence that one business is easier to match to what the person asked for.

The website is only part of the comparison.

When ChatGPT uses web search, you can inspect the sources attached to the answer. OpenAI also says ChatGPT can choose to search the web depending on what the user asks, and that Search can use conversation context, memory and location where relevant.[1][2]

Those source links can tell you where some of the information in the answer is coming from. They can also show you something you wouldn’t see by comparing the 2 homepages side by side.

For example, the competitor is described consistently across its own service pages, professional profiles and other public sources. An industry body, directory, interview or partner page connects that business quite explicitly with the service in question. Or the comparison goes the other way: one source describes a service the competitor no longer offers, another places it in a category far broader than its actual work, and a third gets a basic fact about the business wrong.

The comparison is about the public picture of each business: where it’s described, how consistently, and whether that picture supports the match you keep seeing in the recommendations. The size of the mention pile tells you very little on its own.

This comparison can expose a problem you would miss by looking at your own website alone. A profile created 6 years ago, an old service page or a directory category you forgot existed can all leave a business described in ways that no longer match what it sells now.

A third-party mention still doesn’t prove that it caused the recommendation. It shows you that there is public information connecting the competing business with the topic, which gives you something more specific to investigate.

This is the point at which competitor research needs a fairly firm hand on the steering wheel.

You will almost always find differences between 2 businesses. The question is whether any one of them lines up with the pattern you’re seeing.

For example, the competitor:

In the right context, one of those differences can become relevant. On its own, it doesn’t tell you why ChatGPT recommended that business.

Take Google rankings. If the competitor also ranks strongly for the service in question, that’s something I would look at. But ChatGPT visibility and conventional Google ranking aren’t identical measures, and I’ll deal with that properly in the next article rather than pretending the relationship can be reduced to “they rank higher, therefore ChatGPT chose them”.

The same goes for content volume. A competitor having 150 articles while you have 20 proves that they have more articles. It doesn’t prove that article number 137 tipped ChatGPT over the edge.

This is where copying competitors becomes particularly expensive. If you don’t know which difference matters, copying everything visible is a very efficient way to inherit somebody else’s workload.

What I'm looking for is a difference that keeps making sense as the evidence builds.

For example:

Across all 5 examples, the pattern connects the query, the way the business is described and the public information available about it. The conclusion doesn’t come from one visible feature on the competitor’s website.

That gives you a much narrower problem to investigate.

Before I recommend changing anything, the competitor comparison needs to have done more than make you envious of their website.

I want to see the same difference keep lining up with the pattern across the searches you care about.

That could mean clarifying a service page because the business genuinely isn’t described clearly enough for the need people are searching for. It could mean correcting an old public profile, making a specialist service easier to find, or resolving a mismatch between the way the website describes the business and the categories other public sources use.

And yes, it could mean doing nothing yet.

If the competitor appears in one test, disappears in the next, and another 4 businesses rotate through similar questions, there simply isn’t enough evidence to justify a website project. Finding 27 differences between your site and theirs doesn’t mean any of those differences caused the result.

This is the part of competitor analysis I care about most. You are trying to answer:

You can come away with a concrete fix or another line to investigate. Or you can decide that the evidence still isn’t strong enough to change anything without guessing.

That’s a much stronger position than copying the competitor’s content plan, adding every directory they use and installing whatever piece of schema happens to be visible in their source code.

If you’re happy to investigate it yourself, start by comparing the competitor only against the specific searches where the pattern appears, then look at the descriptions and sources ChatGPT is using. If you’d rather have the query pattern, competing businesses, public sources and your website examined together, that’s part of what I do inan AI Visibility & Accuracy Audit.

What I want to find out is whether the competitor really does look like a clearer, better-supported match for the searches that matter, and whether that gives you a sound reason to change anything on your website or in the public information about your business.

[1] OpenAI Help Center, ChatGPT Search. Current documentation says ranking in ChatGPT Search is based on multiple factors intended to surface reliable, relevant information; there is no way to guarantee top placement. It also covers automatic/manual search, location and OAI-SearchBot access.

[2] OpenAI, Introducing ChatGPT search. Explains that ChatGPT can choose to search based on the question and can use conversation context when bringing in web information.