Skip to content
SEO PlaysAll free plays

How to rank on ChatGPT: get into the sources it reads

AI Search (GEO)effort Mimpact 4/5

ChatGPT holds no opinion about your product. Asked for the best tool in your category, it runs searches, pulls back a set of pages, and recommends what those pages recommend. Your standing is decided by the contents of that set.

The set is smaller and steadier than it looks. Ask the same three buying questions a few times over and the same directories, review sites and listicles keep reappearing — a finite list you can read, count, and work through by hand.

Ahrefs, looking across 75,000 brands, found that mentions of a brand around the web track its ChatGPT visibility far more closely than its backlinks do. So the job is to get named on the pages the model already reads, and several of them will list anyone who fills in the form.

What you get

The pages ChatGPT pulls its answers from in your niche, listed, and your product placed in the ones that take listings or pitches.

Steps

  1. Write down the three prompts a buyer would ask, in their words: "best [category] tools", "[competitor] alternatives", "[category] for [audience]".
  2. Ask each prompt in a fresh ChatGPT session with web search on, and repeat each one three times. Answers vary run to run, and the variation is the measurement.
  3. Copy every cited source into a sheet with four columns: URL, domain, format (listicle, directory, review site, comparison), and how many runs it appeared in.
  4. Mark the recurring sources. Across three prompts and three runs each, any domain cited three or more times is a target.
  5. For each recurring directory or review site, create your listing and fill every field. Most of the big ones cost nothing: in our verified dataset, G2 and Capterra both carry DR 91 and take free listings.
  6. For each recurring listicle, find the author's contact and pitch inclusion with the one concrete fact that fits their list, plus your URL. Two sentences, no adjectives; you are asking for a row, not a review.
  7. Where a prompt has no source you can join, publish your own comparison page in the same format the assistant already cites, honestly including competitors.
  8. Re-run the three prompts every two weeks and log which of your placements the answers now draw on.

How to verify it worked

  • Fresh-session runs of your three prompts: your brand is named in the answer, or a page that lists you appears among the citations.
  • The sheet shows your listing live on at least three of the recurring sources.

Why it works

Being in the retrieved set is what ranking on ChatGPT means. There is no other lever to pull: you cannot edit the model, and the answer gets written after retrieval, out of whatever retrieval brought back.

Evidence

Ahrefs study of 75,000 brands: web mentions correlate 0.664 with ChatGPT visibility, against 0.218 for backlinks — the mention, not the link, is the working currency.

MagicSpace SEO's AI and LLM SEO course (magicspace.agency) documents this exact audit loop and reports brands reaching 20-to-30 percent of answer runs by month one and 50-to-70 percent by month three (course claims, not our measurements).

Last reviewed: 2026-09-01. A play whose evidence stops holding up is rewritten or retired.

Related plays

Browse all free plays

More plays are on the way.

The free plays stay free. The full set becomes a one-time purchase later, not a subscription, and the launch email is the only email you will get.

The launch email is the only one you will get.