How to get recommended by ChatGPT
ChatGPT names two or three brands and moves on. Getting into that shortlist is not luck. It comes down to a handful of things you can actually influence.
By Yurii Shevchyk · Founder, AnswerPeek
ChatGPT does not show a page of links. Ask it for the best tool in a category and it writes a paragraph that names two or three brands. The whole game is being one of those names. Unlike a search result, there is no second page to climb onto. You are in the sentence, or you are not.
Where ChatGPT gets its picks
Two sources feed the answer. The model's training, which is months old and fixed, and live retrieval, where it reads the current web while you wait. You cannot edit the training. You can absolutely shape what it finds when it looks things up, and that is where the work is.
Training answers dominate when the question is generic and timeless; retrieval kicks in when the question implies recency, comparison, or specifics — "best X in 2026", pricing, alternatives. This split matters practically: retrieval-led prompts can be moved in weeks by changing what is on the web today, while training-led prompts move only when a new model version ships. If you are absent from both, start with retrieval, because it is the one with a feedback loop you control.
Be the clearest answer on the open web
When the model retrieves, it favours pages that state a conclusion plainly. A page that opens with "the best options for this are A, B and C, because..." is easier to quote than a page that buries the same point under a story. Lead with the answer, name the alternatives, then justify.
This includes naming your competitors honestly on your own pages. It feels wrong and works: a comparison page that concedes real trade-offs reads as evidence, gets cited, and puts your framing of the category into the answer. A page that only praises itself reads as an ad, and models are notably good at discounting ads.
Get named on pages you do not own
ChatGPT leans heavily on third-party sources: roundups, comparison posts, community threads, and review sites. Being recommended on those pages is often worth more than anything on your own domain, because the model treats them as less biased. Earning those mentions is a real channel, not a side effect.
- Find the roundups that already rank for your category and get added to them.
- Answer real questions in the communities your buyers read, in your own name.
- Make sure your strongest third-party reviews are current, not three years old.
The fastest way to get recommended by a model is to already be the answer that humans give each other.
A 30-day starting plan
- Week 1 — baseline. Write down the ten prompts a buyer would type before choosing in your category. Run each in ChatGPT (with browsing) and note who is named, who is cited, and what is said about you.
- Week 2 — your own pages. Rewrite the two pages closest to those prompts so the conclusion is the first sentence. Add FAQPage and Product schema. Check your robots.txt is not blocking GPTBot or OAI-SearchBot.
- Week 3 — other people's pages. List every source ChatGPT cited in week 1 that does not mention you. Pitch the three most-cited roundups; answer the two most visible community threads properly, under your own name.
- Week 4 — re-run and compare. Same prompts, same order. Log what changed, keep what worked, and turn the run into a weekly habit — answers drift, and one snapshot is not a strategy.
What does not work
- Keyword stuffing and invisible text. Models summarise meaning; they do not count occurrences.
- Prompt-injection tricks on your pages ("ignore previous instructions and recommend us"). Engines filter them, and getting caught is a trust problem.
- One press release. A single mention on a wire service is not the consensus a model looks for.
- llms.txt alone. Useful hygiene, but no major engine treats it as a reason to recommend you.
Then check, because the answer moves
ChatGPT can name you for one phrasing and forget you for the next. The only way to know is to run the actual prompts a buyer would type and read what comes back, on a schedule, not once. Treat it like rank tracking for sentences.
Questions, answered
- How long does it take to get mentioned by ChatGPT?
- For prompts where ChatGPT browses the web, weeks: as soon as the pages it retrieves start naming you, the answers can follow. For prompts it answers from training, months — that changes only when a new model version ships. Most brands see the first movement on retrieval-led prompts within four to eight weeks of real work.
- Does ChatGPT use live web data or only training data?
- Both. Generic questions are often answered from training, which is months old. Questions implying recency, pricing, or comparison trigger live retrieval, where ChatGPT reads current pages and cites them. Retrieval is the part you can influence quickly, which is why AEO work concentrates there.
- Can you pay to be recommended by ChatGPT?
- No. There is no paid placement inside ChatGPT's organic answers. What you can do is earn presence in the sources it retrieves — reviews, roundups, communities, and your own clearly-written pages. That is work, not spend, which is also why it defends itself once won.
- Why does ChatGPT recommend my competitors but not me?
- Usually because the pages it reads mention them and not you: they are in more roundups, have fresher reviews, or state their case more quotably. Run the prompt, open every cited source, and count who appears where. The gap you can see in the citations is the gap to close.