ChatGPT SEO: how to get named when people ask it for a recommendation
What actually happens when ChatGPT answers a buying question, which sources it reads, and the work that changes who it names.
When somebody asks ChatGPT which company they should use, an answer comes back naming two or three. There is no second page. Getting into that answer is the entire objective.
What happens when you ask
Worth understanding in order, because it explains why the usual SEO advice only half works.
It decides whether to search. Questions with current or local intent, and almost all buying questions, trigger a live web search. Questions of general knowledge may not.
It rewrites your question into several searches. Not your exact phrasing. Usually three to six variations.
It reads a handful of results. Roughly five to ten pages, weighted towards pages that compare several options rather than pages that describe one company.
It writes an answer naming a few options, and cites some of what it read.
Two things follow from that, and they are the whole strategy.
The pages that decide the answer are usually not yours. Directories, listicles, review platforms and forum threads. You can have the best page on your own site and be absent, because it was never read.
Being cited and being named are different. ChatGPT can quote your article as a source and then recommend three other companies. Citation is nice. Being named is the result.
First, check you are not blocked
Before optimising anything, open your robots.txt.
Several site builders, security plugins and CDN configurations added blanket blocks on AI user agents over the past two years, often without the site owner being told.
There is an important distinction. Some crawlers collect training data. Another fetches a page live, right now, because a user asked a question. Blocking the first is a legitimate choice. Blocking the second removes you from answers today.
We have audited sites that spent months on content while quietly refusing the crawler at the door. Check this before you conclude you have a content problem.
The work, in order of return
1. Get into the sources it reads
Ask your buyer questions in ChatGPT and read the citations. The sources that recur are your target list.
Then get listed. Directory submissions, a profile on the review platform, an email to whoever wrote the listicle. Unglamorous, cheap, and it is the single highest leverage activity available, because one listing can put you into dozens of answers across many phrasings.
2. Make your facts liftable
ChatGPT summarises text. It cannot summarise an image, a video, or a claim implied by a testimonial.
On your key pages, state plainly in writing: what you do, who for, where, what it costs, how long it takes, and what is genuinely different about you. One sentence each.
A test: paste your homepage text into a blank document and ask a colleague to answer five factual questions about your business using only that. Most homepages fail this.
3. Say the same thing everywhere
Your site, your Google listing, your LinkedIn, every directory. Same description, same categories, same details. Consistency across independent sources is what turns a maybe into a name. Contradictions are a reason to pick somebody else.
4. Reviews
Rating and count come up constantly, especially for local and service answers. Collect them, and state the number on your own site in text so it can be lifted.
5. Publish honest comparisons
The one place your own pages genuinely compete is comparison content. "X vs Y", "alternatives to Z". These match the shape of the question being asked, so they get read.
Write them fairly, including the cases where a competitor is the better choice. A page that says your product wins at everything reads as marketing and gets skipped in favour of one that reads as neutral.
What does not work
Writing for robots. Models read like readers. Text engineered for machines reads badly to both.
Buying guaranteed placement. No vendor controls an organic answer. Anyone selling one is selling something they do not have.
Checking once. Answers vary between runs and phrasings. A single lucky result is not a measurement.
Focusing only on your own blog. Useful for citation. Rarely the reason you get named.
Measuring it
Twenty buyer questions, none containing your brand name, asked in a fresh chat each time with web search on. Record who was named and what was cited. Repeat weekly, and track your competitors in the same run so you can tell a model update from your own decline.
The number that matters is your share of answers, next to theirs. Doing this by hand is about two hours a week; our own Figo runs ten questions across ChatGPT, Claude, Gemini and Perplexity weekly from $49 a month and keeps the history. Either way, the method is what matters, and the method costs nothing.
Questions people ask
Can you rank in ChatGPT?
Not in the sense of a numbered position. You are either named in an answer or you are not. That is the whole scoreboard, which is why it is measured as a share across many questions rather than a rank.
Does ChatGPT read my website?
Sometimes, if your page comes back in one of the searches it runs. More often the deciding sources are third party lists, directories and review sites, because they cover several options at once.
Should I let OpenAI's crawlers access my site?
If you want to appear in answers, yes for the crawler that fetches pages to answer live questions. Whether to allow training crawlers is a separate business decision.
How long does it take to see a change?
Typically four to eight weeks, mostly determined by how quickly the third party sources you got listed on are re crawled.
See it on your own competitors
Figo checks their ads, pages, rankings, reviews and AI answers every week, then tells you what to do in plain words. Set up in two minutes.
Keep reading