How to get your business recommended by ChatGPT
You can’t buy your way into a ChatGPT recommendation, and you can’t trick your way in for long. But the work that earns mentions is knowable — six moves, in order of leverage.
When ChatGPT recommends providers, it assembles the answer from evidence — what credible sources say about who exists, who serves the area, and who is well regarded. It cross-references directories, review platforms, maps data and articles, and it names the businesses those sources agree on. That means the path to being named runs through work you control, and none of it requires permission from OpenAI.
Two ground rules before the steps. First, there is no shortcut: engines weigh corroboration, so a claim that exists in only one place — your own website — barely counts. Second, never judge progress by a single chat: the same question returns different names run to run, so only frequencies across repeated runs mean anything (the evidence). With that said — six moves, in order of leverage.
Find out where you stand
Before optimizing anything, get a baseline. Ask the engines your buyers’ questions — repeatedly, because single answers vary — and record who gets named. If you’d rather not run dozens of chats by hand, a free scan does it across six engines and stores every answer verbatim.
Get present in the sources engines cite
AI answers lean on directories, review platforms, maps data and articles the engine treats as credible. The citations in your category’s answers tell you which sources matter for you — that list is the target, not “everywhere.”
Make your NAP boringly consistent
Name, address, phone, category — identical on every listing. Engines cross-check sources; disagreement reads as risk, and risky names get skipped in favor of safe ones.
Earn words from third parties
Engines weight what others say about you far above your own copy. Steady reviews that name your service and city, local press, supplier and partner pages, industry directories — each is another independent witness that you’re real and good.
Publish quotable answers
Write pages that answer buyer questions in plain sentences an engine could safely repeat: who you serve, where, what it costs, how fast you respond. Add FAQ content and structured data so the facts are unambiguous to machines as well as people.
Re-measure, and keep what moves the number
AI answers vary run to run, so progress only shows up in frequencies over repeated runs — the same discipline behind our leaderboard. Measure monthly with the same questions and treat the mention rate as the KPI.
Why this order
The sequence follows the evidence chain. Measurement comes first because everything after it is guesswork without a baseline — you can’t tell whether the work moved anything. Cited-source presence comes before content because engines can only repeat what their trusted sources corroborate; the most quotable website in the world doesn’t help if the engine never consults it about you. Consistency comes before promotion because new attention amplifies whatever the sources say — including the contradictions. Reviews and third-party words compound everything upstream. Content converts the credibility you’ve built into sentences an engine can actually use.
How long until it works
Honestly: no fixed timeline exists, and anyone quoting one is guessing. Retrieval-backed surfaces can reflect a completed, corroborated profile within weeks; trained knowledge refreshes on the engines’ own schedules. What you control is the cadence of evidence and the cadence of measurement. Businesses that do the work and re-measure monthly know exactly when it lands; businesses that don’t are guessing in both directions.
What not to waste time on
Keyword-stuffed pages written for machines, mass directory spam, and anyone selling a guaranteed placement. There is no placement to sell — engines assemble answers from evidence, and shortcuts read as noise. The boring fundamentals above are what the evidence is made of.
Reading your baseline like an operator
A good baseline tells you three things, and each maps to work. Your mention rate per engine shows where you already have a foothold and where you’re invisible — surfaces behave differently, so a strong ChatGPT number doesn’t transfer to Gemini automatically. The names appearing instead of you define the real competitive set: often it isn’t who you compete with on price, it’s who out-documents you. And the cited sources are the to-do list — every citation pointing at a profile you haven’t claimed is leverage sitting on the table.
One more move if you’re in a category we already measure: find your business on the leaderboard and claim the listing. Claiming is free, never changes your number, and gives you a verified badge, the ability to respond, and a standing view of your movement month over month.
No. Organic AI answers carry no paid placement. Anyone selling a guaranteed AI recommendation is selling something they don’t control.
Answers are generated fresh each run, so the same question returns different names day to day. That’s why the honest metric is how often you appear across many runs — not whether you appeared once.
Yes. The assistants consult overlapping evidence, so the work above moves all of them at different speeds — which is why we measure six engines, not one.
Want the baseline without the manual work? The free scan runs your buyers’ questions across six engines, repeatedly, and hands you the count — plus who the engines named instead of you and the sources they leaned on.