How to Check Whether ChatGPT Recommends Your Product (And Get Into the Sources It Cites)

Your buyers have a new first step. Before they open Google, before they read a comparison post, a growing share of them type the question straight into ChatGPT or Perplexity: "what is the best X for Y". The assistant returns a short list of named products. If you are not on that list, you were never in the running, and you will never see it in your analytics because there was no click to miss.

This is measurable, and it is worth measuring. Here is how to test where you stand, why the answer is usually disappointing, and what actually moves it.

Test It Properly, Not Flatteringly

The mistake is asking the assistant about yourself. "What do you know about Acme Analytics" prompts the model to talk about Acme Analytics. It tells you nothing about whether it would ever recommend you.

Ask the buying question instead, exactly as a customer would phrase it, in a fresh conversation with no history:

• "Best [category] for [audience]"
• "[Category] alternatives to [the market leader]"
• "Who should I use for [job to be done]"
• "Cheapest [category] that still [key requirement]"
• "Is [competitor] worth it, what else should I look at"

Run each one in ChatGPT, Perplexity and Google AI Overviews. Perplexity is the most useful of the three for diagnosis, because it shows its sources. Those citations are the real prize: they tell you precisely which pages a model reads when it answers your category's buying question.

Record two things per query: whether you were named, and which sources were cited. A simple table with five questions, three assistants and a date is enough. Check it monthly, not daily. These answers shift on their own schedule and daily checking only produces noise and frustration.

Why You Are Probably Not There

When we ran this test on our own category, a well-known assistant named twenty providers of done-for-you directory submissions. We were not one of them, despite explicitly allowing GPTBot, ClaudeBot and PerplexityBot in our robots file. That was a useful thing to learn about ourselves, and the reason is not mysterious.

Being crawlable is a floor, not a lever. Allowing the AI crawlers removes a blocker. It does not give a model a reason to name you. Plenty of sites allow every crawler and never appear.

Models favour claims they can find more than once. A product that exists only on its own website is a single unverified source. A product with twenty directory listings, reviews on G2 and Capterra, and a few independent mentions looks like an established option, because the same facts appear in places that are not you.

They favour claims they can compare. Look at the products that do get named and you will notice their promises are structured and checkable: a named package, a price as a number, a scope as a list. "30+ live listings for $149" is quotable. "Grow your online presence" is not. If your value is genuinely different, state it in a form that survives being pulled into a comparison table.

The Three Things That Actually Move It

1. Get listed where the sources are

The large directories are not just backlink sources, they are cited sources. G2, Capterra, Product Hunt, AlternativeTo and SaaSHub show up constantly in the citation lists of AI answers, because they hold structured, comparable, frequently updated data about products in a category.

A listing puts your product inside a page the model already trusts, described in the same format as your competitors. This is the same work as directory submission for SEO, aimed at a different reader. If you want the shortcut, we do this as a done-for-you service, and if you would rather start yourself, our list of free directories worth submitting to is a reasonable first hour.

2. Structure your own pages to be quotable

Assistants lift answers out of pages that are already shaped like answers. In practice that means:

• A real FAQ section with the question written the way people ask it, marked up as FAQPage structured data so the question and answer pair is machine-readable.
• Pricing stated as numbers on the page, not hidden behind a form. A model cannot quote a price it cannot see.
• Scope written as a list. What is included, what is not, how long it takes.
• One clear sentence that says what the product does, in the vocabulary buyers use rather than your internal naming.

None of this is a trick. It is the same clarity that helps a human decide, which is why it works.

3. Earn mentions that are not yours

The hardest and most durable of the three. Listicles in your category, comparison posts, threads on Reddit and Hacker News, community roundups. Repetition across independent sources is the signal that separates a product a model will name from one it will not.

You cannot buy this credibly and you should not try. What you can do is make yourself easy to include: publish the numbers, the pricing and the limitations plainly so that anyone writing a roundup can use you without doing extra research.

What Nobody Can Promise You

Be sceptical of anyone selling guaranteed AI visibility. Nobody outside the labs knows the retrieval and training schedules, the answers vary between users and sessions, and the same question can produce a different list a week later. There is no equivalent of a rank tracker with an authoritative number behind it.

What you can honestly do is improve the inputs the models demonstrably use, then measure your presence in the output over months. That is a slower promise than most of this category makes, and it is the only one that holds up.

The Practical Version

If you want the short version: run five buying questions across three assistants today and write down the result, including which sources were cited. Get listed in the directories that appear in those citations. Put a real FAQ and a real price on your own pages. Then check again in a month.

The first step costs twenty minutes and tells you where you stand. Most founders have never run it, which is exactly why the answer is usually a surprise.

If you would rather not spend your evenings filling in submission forms, we submit your product by hand to the directories that matter and send you proof of every single listing.