Matt Turley, Continuum. Updated October 6, 2026.
When I asked ChatGPT for the best realtor in Phoenix and Dallas, it cited one rankings site in 11 of 12 answers. Being on the sites it reads is the lever, and your own page is the second one.
Be listed on the sites ChatGPT already reads for your kind of work, and have one page that answers the buyer's exact question in its first hundred words. A Google ranking alone will not get you named: a page can sit in the top 10 of a web search and still be left out of the AI answer.
Start with what ChatGPT actually reads. On September 28, 2026 I asked it for the best real estate agent in Phoenix and in Dallas, 12 answers in all, and it cited one rankings site, therealtorrankings.com, in 11 of the 12. Being found and being named are two separate steps. In an August 2026 study of 45 queries where a site ranked at position 12 or better on Google, a live web-search index surfaced the site on 27 of them, 60 percent. A browsing AI model cited it on only 7, 15.6 percent. Ranking bought retrieval. It did not buy the citation.
A citation is still not a mention. The AI can pull your page in as a source and never say your name. Being cited as a link is a smaller win than being named. This is consistent with a November 2023 paper on generative engine optimization (arXiv 2311.09735), which found that pages built with citations and statistics see up to a 40 percent lift in generative-engine visibility, a different lever from a Google ranking.
Mostly other people's sites. On September 28, 2026 I asked ChatGPT and Gemini who the best real estate agent is in Phoenix and in Dallas, 12 answers per tool. ChatGPT cited therealtorrankings.com in 11 of 12 answers. Gemini cited fastexpert.com in 11 of 12. Both leaned hardest on rankings, agent-matching and local magazine sites, and the two tools mostly read different sites. It is a small sample, two cities on one day and 24 answers in total.
Bookkeepers look the same. In September 2026 I checked bookkeepers in Denver with 10 buyer questions, each asked three times on ChatGPT, Perplexity and Gemini. The sources cited most were directories and review sites. Clutch.co was cited in 66 of 90 answers and expertise.com in 60 of 90. Sound Advice Bookkeeping, the bookkeeper named most, came up on 19 of 30 checks and on all three tools.
Which sites matter depends on the field. On 12 Charleston trip questions the engines cited 302 links, and TripAdvisor was 47 of them while Reddit was 4. For software questions it flips, and Reddit, GitHub and dev.to lead.
For the bigger picture, I pooled a week of these checks, 3,118 answers and 39,763 citations across ChatGPT, Perplexity and Gemini, into one study: which sources AI answers cite when people ask who to hire.
So start by asking the engines your buyers' questions and writing down the sites they cite. Then get listed and reviewed on those sites. That is usually the bigger lever, and your own page is the second one.
In that August study the pages that won the citation had a few habits in common. I write my own answer pages this way now.
Here is the order I would work in, using only what is covered on this page.
No. ChatGPT can cite your page as a source and still name someone else in the answer. Count the two apart. A name in the answer is the win that sends buyers to you. A citation alone is smaller.
Check this before you write anything. Perplexity's own bot documentation recommends allowing PerplexityBot in robots.txt, and says a web application firewall may need to whitelist its bots. Changes to robots.txt can take up to 24 hours to show up on their side.
OpenAI's crawler documentation names OAI-SearchBot as the one that surfaces sites in the live search results ChatGPT cites, separate from GPTBot, which trains its models. Allowing OAI-SearchBot is what keeps you eligible for citation; blocking GPTBot does not affect it.
Do the same review for every AI crawler you care about. A good page the crawler cannot read will never be quoted.
Write down the questions a real buyer would ask. Ask each one three times, because answers change from run to run, and only count a question as won when you show up in most of the runs.
Keep named and cited apart, and keep each engine apart too. One engine can start naming you while another sits still, and an engine can link your site as a source while recommending someone else. Being cited and being named are two different wins, so track both, per engine.
If you would rather not run the questions yourself, my AI Answer Check does it across ChatGPT, Perplexity and Gemini and shows who they name instead of you, plus the fixes worth the most. It is $99, and it does not promise what AI tools will say.
Get Users is the monthly version: the Proving rate is $500 a month while we prove the method works for you, then the Founding Client rate, $999 a month locked for life once the written goal is hit (standard $1,999 a month), month to month. We do the work and you approve each week’s batch, or you make the changes from our list at the same price, with a weekly scoreboard of what the three tools say about you, and there are no ads and no free trial.