Ask any marketer how to get your company mentioned in ChatGPT and you get the same list back. Build entity authority. Get on G2 and Capterra. Earn third-party mentions. Add schema. Collect reviews. Show up on Reddit. Do not block the crawler. None of it is wrong. All of it is on the same eight blog posts, which is why it no longer separates anyone.
The problem is not the list. The problem is that it treats “ChatGPT” as one thing, and it is two.
Memory and retrieval are different machines
Default ChatGPT answers a lot of questions from memory. The model was trained on a snapshot of the web that closed months ago, and for a question it feels confident about, it answers straight from that frozen snapshot. No search happens. Nothing on the live web is consulted.
Sometimes ChatGPT searches instead. It sends a crawler out, fetches live pages, and builds the answer from what it just read. You have seen this when it shows a “Searching the web” line or drops citations under the answer.
These are two different machines wearing one logo, and the moves that get you into each are opposite.
To show up when ChatGPT retrieves live, you need to be crawlable, be the clear liftable answer to the question, and be present in the handful of sources it actually fetches. All of that you can influence inside a crawl cycle, measured in weeks.
To show up when ChatGPT answers from memory, you need to have been described so consistently, across enough of the trusted web, that the next training run learns you as the answer. You cannot schedule that. It lands on OpenAI’s training calendar, months out, if it lands at all.
So the honest response to “how do I get mentioned in ChatGPT” is a question back: which ChatGPT, and are you prepared to wait on the one you cannot rush? Every checklist skips this, then quietly folds tactics for both surfaces into one to-do list, so you cannot tell which of your effort pays off this quarter and which is a bet on a training run you do not control.
The ten-minute test that sorts your effort
Before you touch the checklist, open ChatGPT and ask it your real buyer questions. The ones a prospect would actually type: best tool for X, alternative to Y, how do teams like mine solve Z.
Watch what it does. If it searches, shows sources, cites pages, that question runs on retrieval, and you can move it in weeks by working the sources it pulls. If it answers cold, no search, no citations, that question runs on memory, and no single thing you publish this month will change the answer. It updates on a training refresh, on its own timeline, a slower clock than most guides admit.
Ten minutes of this tells you more than any “how to appear in ChatGPT” guide, because it tells you which surface your category actually lives on. Some categories get searched almost every time. Some get answered from memory almost every time. The tactics read the same on paper and are completely different bets depending on which one you are looking at.
You do not get mentioned in ChatGPT. You get mentioned in its sources.
Here is the part the checklists get right for the wrong reason. There is no ChatGPT profile to claim, no submission form, no field you fill in. There is nothing you add to your own website that inserts you into the model. Everything that moves ChatGPT lives on other people’s pages: the review sites it was trained on or retrieves from, the comparison posts, the community threads, the analyst mentions.
That is why the list points you at G2 and third-party mentions. It is right about the where. It is wrong about the framing. It reads like “optimize your site for ChatGPT,” when the truthful version is “get the rest of the web to describe you the way you want to be found,” because ChatGPT is repeating them, not you. The single page that does the least work here is your own homepage. A claim you make about yourself is the weakest signal on the web. The same claim, made by sources ChatGPT already trusts, is the whole game.
What your own page is actually for
Your site still matters, in exactly two narrow ways, both tied to the retrieval surface.
First, be liftable. When ChatGPT does fetch live, it quotes the page that answers the question cleanly and verifiably in a few sentences, not the page that buries the answer under a narrative. A clear, self-contained, specific answer near the top is what gets pulled.
Second, do not accidentally lock the door. The retrieval crawler has to be able to reach you. Plenty of companies copy a “block all AI bots” rule into robots.txt and quietly shut out the exact crawler that decides whether they exist in the searched answer. Getting this right does not get you recommended. It gets you eligible. The third-party consensus is what gets you chosen.
What I got wrong
I used to run “get us into ChatGPT” as one project with a weekly status update. For the first month, the update said nothing was moving, and the client started to think the program had failed.
What was actually happening: the retrieval-mode wins showed up in two to three weeks, on the questions where ChatGPT searched. But the client kept testing the default answer, the cold one you get when you just ask without it searching, because that is the first thing everyone types. That answer did not budge, because it only updates on a training refresh I have no button for. I was measuring a slow, unmovable surface with a weekly ruler and calling the whole thing dead.
Now I split the two on day one. Run the questions, label each as searched or answered-from-memory, pour the work into the retrieval surface because it is the one that moves, and say out loud, before anyone panics, that the default memory answer lags by a training cycle. Same tactics as everyone’s checklist. Completely different expectations, which is the part that actually decides whether the program survives month one.
Where this leaves you
If you take one thing from this: stop asking how to get into ChatGPT and start asking which ChatGPT your buyers are using for your category, and which of your moves can move it before the next training run. Run the questions. Work the sources, not your homepage. Be liftable and be reachable. Then judge the program on the surface you can actually move, not the one that updates on someone else’s schedule.
That is the difference between doing the checklist and knowing why half of it will not pay off for months.
If you want this run on your category, the buyer questions mapped and the sources worked, that is the engagement, and it runs on a documented method.