How does ChatGPT decide which brands to recommend?

For 'best X' questions it often reads a few roundup and review pages and recommends the brands on them. Search Engine Land found the share of category listicles a brand appears in correlates r=0.64 with ChatGPT naming it.

Why does ChatGPT recommend me when I ask about my brand, but not otherwise?

Brand questions are easy. In our study of 5 small SEO agencies, they were recommended in 114 of 130 answers when asked by name, but in only 14 of 1,119 answers to category questions.

Does llms.txt or schema markup help?

Not on their own. An SEJ study found no link between llms.txt and AI citations across 300K domains, and Ahrefs saw no significant citation lift from schema in ChatGPT or AI Mode.

How long until changes show up?

It varies by assistant. Search Engine Land found Perplexity picks up new pages within 1 to 3 days on uncontested queries. Re-scan at 30 days for an early read and at 60 days to compare.

Guides · Updated Oct 2, 2026 · By Rajan Chavada, founder of Soarce

How to get recommended by ChatGPT, Gemini and Perplexity

How to get recommended by ChatGPT, Gemini and Perplexity: find the lists they read, get onto them, fix your own pages, and re-scan to see what changed.

The short answer

ChatGPT, Gemini and Perplexity mostly recommend the brands that appear on the lists they read: roundups, reviews and comparison pages. To get recommended, find the lists they cite for your buyers' questions, get onto them, and keep your own pages clear and current. Then ask the same questions again to see what changed.

Category questions matter more than brand questions

A buyer who already knows your name will ask about you, and the assistant will usually say good things. The hard part is the question before that: "what's the best X for Y?" That's where buyers find new options, and that's where small brands are missing.

We saw this in our own study of 5 small SEO agencies that sell AI search work. On "best agency" questions they were recommended in 14 of 1,119 answers (1.3%). When the question named the agency, they were recommended in 114 of 130 (88%).

Question type	Recommended	Share
Category ("best SEO agency for...")	14 of 1,119	1.3%
Brand (asked by name)	114 of 130	88%

If you only test your own name, you'll think you're fine. Full method in the agency study.

Assistants pick from the lists they read

When an assistant answers a "best X" question, it often searches the web and leans on a few roundup pages. Search Engine Land found the share of category listicles a brand appears in correlates r=0.64 with ChatGPT naming it (Apr 2026). Peec found that being #1 inside a cited listicle moves a brand 0.8 to 1.8 positions earlier in the answer (200K responses).

Your own list doesn't help much. Lily Ray found that when a brand's own "best X" listicle was cited, the brand was left out of the recommendations 69% of the time (100 B2B queries). Mentions on other sites count for more: Ahrefs found web mentions correlate 0.66 with AI Overview mentions, against 0.22 for backlinks (75K brands, May 2026).

5 steps

Scan. Ask your buyers' real questions on each assistant, many times. Answers vary a lot: SparkToro found the odds of ChatGPT giving the same brand list twice are under 1 in 100.
Find the cited lists. Note the pages each assistant cited when it picked someone else. Rank them by how many of your rivals they name.
Pitch the editors. Email the person who updates each list. Give them a reason to add you: a fact, a use case the list misses, a free account to test.
Fix your own pages. Say plainly what you do, who it's for and what it costs. Keep pages current: AirOps found 70% of cited pages were updated within the past year. Reviews help too: Seer Interactive found brands with even 1 to 13 Trustpilot reviews went from 1% to 53.5% citation rate (804K responses, May 2026).
Re-scan. Ask the same questions the same number of times at 30 days for an early read, and again at 60 days to compare. Compare the share per assistant and check whether the lists you pitched now name you.

Two things that don't help on their own: llms.txt (an SEJ study found no link to AI citations across 300K domains) and schema markup (Ahrefs saw no significant citation lift in ChatGPT or AI Mode).

Our own baseline: 0 of 224

We run Soarce, a tool that does steps 1 and 2 for you. On Oct 2, 2026, we scanned our own site. Soarce was recommended in 0 of 224 answers: ChatGPT 0 of 86, and Google AI Mode, Gemini and Perplexity 0 of 46 each.

The assistants recommended Profound (109), Peec AI (75), Semrush (73), Otterly.AI (70), Ahrefs Brand Radar (44), Scrunch AI (35) and SE Ranking (24). They cited 10 lists, including zapier.com, explodingtopics.com and builtin.com. Profound was on 8 of the 10 and #1 on 4.

So we're doing the 5 steps on ourselves: pitching those 10 lists, writing pages like this one, and re-scanning to show the result. Follow along in the build log. For a tool comparison, see AI visibility tools for small marketing teams.

Questions

How does ChatGPT decide which brands to recommend?: For 'best X' questions it often reads a few roundup and review pages and recommends the brands on them. Search Engine Land found the share of category listicles a brand appears in correlates r=0.64 with ChatGPT naming it.
Why does ChatGPT recommend me when I ask about my brand, but not otherwise?: Brand questions are easy. In our study of 5 small SEO agencies, they were recommended in 114 of 130 answers when asked by name, but in only 14 of 1,119 answers to category questions.
Does llms.txt or schema markup help?: Not on their own. An SEJ study found no link between llms.txt and AI citations across 300K domains, and Ahrefs saw no significant citation lift from schema in ChatGPT or AI Mode.
How long until changes show up?: It varies by assistant. Search Engine Land found Perplexity picks up new pages within 1 to 3 days on uncontested queries. Re-scan at 30 days for an early read and at 60 days to compare.

Run a free scan