ChatGPT serves 800 million people a week,1 and a growing share of them ask it buying questions: who to hire, what to buy, which tool to pick. When it answers, it names a handful of brands. This is the practical, no-hacks playbook for becoming one of them.
First, the mechanics that make everything else make sense. ChatGPT answers from two places: live web search (it rewrites your question into queries and reads the results — citations increasingly come from outside the top ranked results — top-10 pages now account for just 38% of citations, down from ~76% a year ago2) and training memory (Semrush found roughly two-thirds of prompts never touch the live web3). You influence the first by being present on the pages it retrieves today, and the second by being consistently written about long enough to be part of what it remembers. Same work, two payoff horizons.
List 10–25 questions your buyers would ask right before choosing someone in your category — “best X for Y,” “X vs Y,” “X alternatives,” “who should I hire for Z.” Pull phrasing from sales calls and support tickets, not from keyword tools — buyers phrase things differently than marketers. (Full method: how to choose prompts to track.)
Ask ChatGPT each question. Record every brand named, in order. That list — not your traditional competitor set — is who you're actually up against. Expect surprises: brands you've never considered rivals show up constantly, because AI's consideration set is built from the web's evidence, not your market map. One run is noisy (answers vary meaningfully run-to-run4), so ask each question a few times or track on a cadence before trusting the pattern.
For every question where competitors show and you don't, dig into why: check the citations when search fires, and ask follow-ups (“why those three? what sources rank them highly?”). You'll converge on a short list of specific pages — a couple of “best of” listicles, one or two review platforms, a Reddit thread or three. That short list is your entire roadmap. Ahrefs' 17-million-citation analysis confirms the pattern: a concentrated set of trusted pages supplies most of the evidence.5
Where the evidence behind AI answers lives
Share of 17,000+ citations by source type · Beacon production tracking, 30 days to Jul 17, 2026
Half the evidence lives on pages no brand controls — which is why steps 3 and 4 are the heart of this playbook.
Now do the unglamorous work, in priority order:
Listicles:pitch the authors of the specific “best X” articles ChatGPT keeps reading. Give them a reason to add you — real differentiators, real numbers, a quote. One placement on a list AI trusts echoes through thousands of answers.
Review platforms:build genuine review volume wherever your category's citations point (G2 for SaaS, Yelp/Google for local, category sites elsewhere). Volume and recency both matter; fake anything and you're on the wrong side of an FTC rule that now explicitly bans it.6
Communities: answer the actual Reddit and forum threads in the citations — as yourself, disclosed, genuinely useful. Authentic participation reads as evidence; astroturfing reads as risk.
For each tracked question, publish one page that answers it directly — claim in the first paragraph, evidence after, phrased the way a model would want to quote it. Research on generative engines found that adding quotable statistics, citations, and authoritative phrasing measurably lifts visibility in AI answers, while keyword stuffing reduces it.7You're not writing for dwell time; you're writing to be lifted verbatim.
Re-run your question set weekly or daily. When a mention appears, trace which move earned it — new listicle placement, review milestone, a thread that took off — and double down. The loop compounds: today's answers get cited and re-ingested, so every mention you earn makes the next one more likely. Expect the first movement in weeks on search-grounded questions, longer on memory-answered ones. The brands that win treat this as a program, not a project.
Beacon automates the measurement half of this playbook: it runs your buyers' questions across ChatGPT and four other AI models every day, extracts the sources behind each answer, and shows which of your moves earned each new mention — so steps 2, 3, and 6 happen while you focus on step 4, the part only a human can do.
Beacon runs your buyers' questions across ChatGPT and four other AI models daily — who gets named, which sources decide it, and what to do next.