Ask ChatGPT for “the best anything” and it answers with a short, confident list of names. If yours isn't on it, that list is quietly costing you customers — and no, there is no form to fill in and no ad to buy. This guide covers what actually moves the answer. It comes from measurement, not theory: Orbator's AI Recommendation Index re-asks thousands of buyer questions across ChatGPT, Claude, Gemini, Perplexity and Grok every week and records which sources each engine read to decide — so every step below is something we've watched work in the data, and every number on this page is measured, not invented.
Yes — but not by asking OpenAI. ChatGPT builds its recommendations by reading sources it trusts — review platforms, rankings, editorial articles, community threads, your own website — and repeating what they collectively say. Change what those sources say about you and you change the answer. The work, in order: see what ChatGPT recommends in your category today, find the sources behind that answer, get present and well-described on them, publish the missing answer yourself, and re-check until you're on the list. The first step is free: run a free AI visibility check and see where you stand across all five major engines in about twenty seconds.
Ask an AI about this page
Before changing the answer, read it: ask ChatGPT the question your customer asks, in their words, and write down who gets named.
Open ChatGPT and ask the way a real buyer would — not your industry’s label for what you do, but the words a customer uses when they have the problem. “Best project management tool for a small agency.” “Who should I use for payroll?” Try three or four phrasings, because the engines answer occasions, not companies — you can be the answer for one phrasing and absent from its neighbor.
Record three things: who is named, in what order, and how you are described if you appear at all. That snapshot is your baseline — everything else in this guide is measured against it. If you'd rather not do this by hand, the free scan asks the buyer questions for your category across ChatGPT, Claude, Gemini, Perplexity and Grok at once and shows the verdict per engine, which matters because the engines frequently disagree — the live disagreement rate is published on our data page.
ChatGPT's recommendations trace back to readable sources — and they are usually not the sources you’d guess.
Every recommendation has a paper trail. When ChatGPT browses, it cites pages; when it answers from training, it is repeating what widely-read pages said. Either way, a specific set of sources decides your category’s answer — and finding that set is the single highest-leverage move in this guide, because it turns “get famous on the internet” into a finite to-do list.
Expect surprises. In the software categories we measure, the citation graph is dominated by blogs, rankings and editorial pages — not the big review platforms. G2, for instance, accounted for just 552 citations across every category we track in a recent 28-day window. And as of July 2026, 45.5% of all AI citations in our corpus point at domains owned by vendors in the category — the industry writing about itself (live figure on /data). Both facts are good news for you: the sources that decide the answer are pages anyone can appear on, or write.
Claim, complete and keep current every profile and listing that shows up in your category's sources — starting with the ones every engine reads.
Work down your source list and split it in two: places you can simply join (directories, review platforms, listing sites, comparison databases) and places you have to earn (editorial articles, rankings, press). Do the joinable half first — it is fast, mostly free, and entirely under your control. A half-empty profile with no description, wrong category and no reviews teaches the engines nothing; a complete one hands them the exact sentences they repeat.
Prioritize the overlap. In our citation data, the most valuable surfaces are the ones multiple engines read — some sources in a category are cited by four of the five engines we track, and being described well there moves every answer at once. And don’t skip the platforms you think of as social: when we analyzed our own category’s citation graph in July 2026, LinkedIn and YouTube were both among the top five cited sources.
Pitch the rankings, roundups and articles that already decide your category’s answer; independent coverage stands out in a graph full of vendor marketing.
Now the earned half of the list. Find the “best X” roundups, comparison articles and industry blogs that appear behind your category’s answers, and pitch the ones that don’t mention you. Be specific and easy to publish: who you’re for, what you actually do, real pricing, and one thing that is genuinely different. Writers keep lists like these updated because freshness is how the lists themselves stay cited — a concrete, honest pitch is doing them a favor.
Here’s why this step punches above its weight: with 45.5% of AI citations pointing at vendor-owned domains, the engines are drowning in companies describing themselves. A third party describing you is the scarcer signal. One well-placed mention in an independent ranking that several engines read can do more than a dozen pages on your own site.
Three page formats win AI citations over and over: ranked lists, how-to guides, and pages of real data. Publish the one your category is missing.
When we pulled apart which pages actually earn citations in our corpus, nearly all of them were one of three formats: ranked comparisons (“best X for Y”), how-to guides answering one specific question, and pages of real data or research. Not homepages, not feature tours — pages that answer a buyer’s question completely enough that an engine can quote them.
You don’t need to be a publisher to do this. Write the page a buyer would want: say who each option is for (including competitors — that’s what makes it credible), quote real prices, answer the obvious questions plainly, and put a visible “Updated” date on it — the winning pages we tore down keep their stamps current, and stale pages fall out of answers. One honest, complete page beats ten thin ones.
AI answers change as the engines re-read their sources; measure monthly so you know what worked and catch it when you fall out.
This is the step everyone skips. ChatGPT’s answers are not fixed — they shift as models update and sources change, which means the work above pays out gradually and can also quietly unwind. Ask your baseline questions again every month: are you named now, and how are you described? Did the source you pitched show up behind the answer?
If checking five engines by hand every month sounds like a chore, that is the product we built: start with the free check, and Orbator's weekly tracking re-asks your category's questions across all five engines and shows which sources each one read — so you see the answer change, and know why. For picking a tracking tool (ours or anyone's), our honest comparison for small and local businesses is here.
There is no fixed timeline and anyone promising one is guessing. AI answers shift as engines re-read their sources — in our weekly Index measurements, category answers change week to week. Profile and listing fixes are read on the next crawl; earned editorial mentions take as long as the pitch takes. Plan in months, measure monthly, and treat any single re-check as one sample, not a verdict.
No. There is no paid placement in ChatGPT’s recommendations — the answers are assembled from sources the model reads. That cuts both ways: you can’t buy your way in, and neither can your competitors. The lever that exists is changing what the sources say, which is exactly what this guide covers.
Because the sources it reads describe them for that question and not you. In our citation data, 45.5% of cited domains are vendor-owned — often it is literally the competitor’s own comparison pages and guides being read. It is rarely about who has the better product; it is about who the readable sources say is the answer.
Largely, yes — every major engine builds recommendations by reading sources, and the most valuable sources in our data are the ones several engines read at once. But the engines genuinely disagree with each other (the live agreement rate is on our /data page), which is why we measure all five instead of treating ChatGPT as the whole game.
Two, and you should use both. Ask ChatGPT yourself, phrased the way your customer would ask — it costs nothing and takes a minute. And Orbator’s free scan asks the buyer questions for your category across ChatGPT, Claude, Gemini, Perplexity and Grok, no card required, and shows which sources each engine read to decide.
Numbers on this page: Orbator AI Recommendation Index citation corpus, 28-day window, July 2026 (45.5% vendor-owned share of 103,733 citations; 552 G2 citations; top cited sources and formats). Live figures, sample sizes and the CC BY 4.0 license at orbator.io/data. Where we don't have a number, none is given.
Find out what ChatGPT says about your category — before your next customer does.
Check Your AI Visibility Free →© 2026 Orbator. All rights reserved.