How to show up in ChatGPT, Perplexity and Gemini answers

July 14, 2026 · 9 min · Spike research

Your buyers stopped starting at Google. They open ChatGPT, Perplexity, or Gemini, type "best tool for X," and act on whatever the model names first. If you want to show up in ChatGPT, Perplexity and Gemini answers — and get recommended by AI at the exact moment someone is ready to choose — you need a different playbook than classic SEO. AI answers are not ten blue links you can climb one rank at a time. They are a single confident paragraph, and either your name is in it or you do not exist. This guide breaks down how the engines decide, the seven things that actually get you cited, and a 30-day plan to move the needle.

How ChatGPT, Perplexity and Gemini pick what to recommend

Every AI answer is assembled two ways, and the mix depends on the engine.

Retrieval-first engines run a live web search behind your question, pull a handful of pages, and write an answer grounded in what they just read — with citations. Perplexity works this way by default. ChatGPT does it whenever search is on, which is most of the time now. For these, your job is simple to state and hard to do: be one of the pages they retrieve, and be quotable enough that the model lifts your sentence into its answer.

Memory-plus-retrieval engines like Gemini lean partly on what they absorbed during training and partly on fresh retrieval from Google's index. That means two things get you named: being well-represented across the web at training time (so the model already "knows" you), and being retrievable and current right now.

The practical takeaway: retrieval-first engines reward fresh, crawlable, well-cited pages you can influence this quarter. Memory-blended engines also reward long-term reputation you build over months. You optimize for both by making your site machine-readable and by earning mentions everywhere buyers and models look. If you want the deeper mechanics, read how AI assistants decide and our Answer Engine Optimization guide.

The 7 things that get you cited

This is the core playbook. Marketers and engineers can execute all of it. Each step maps to one of the 7 fixes that make AI cite you, and each works for a concrete reason.

1. Make sure AI crawlers can actually read your site. Most AI crawlers fetch your raw HTML and do not run JavaScript. If your content only appears after a client-side render, the bot sees an empty shell and cannot quote what it cannot see. Serve real HTML through server-side rendering or static generation, and check your key pages by viewing source, not the live DOM. This is the single most common reason a good product is invisible to AI — see why AI can't see your site.

2. Publish an llms.txt file. This is a plain-text map at your root that points models to your most important, most quotable pages — docs, pricing, comparisons, FAQs. It works because it hands the model a clean summary of what you offer without making it crawl and guess. Think of it as a menu written for machines. Our complete llms.txt guide shows the exact format.

3. Explicitly allow AI crawlers in robots.txt. GPTBot, PerplexityBot, Google-Extended, ClaudeBot, and others each honor robots.txt. If you block them — or copied a restrictive default — you have opted out of being cited entirely. Add allow rules for the agents you want, then confirm they are hitting your pages in your server logs. You cannot be recommended by an engine you forbade from reading you.

4. Add schema and JSON-LD structured data. Markup like Organization, Product, FAQPage, and Review tells the model what your page means, not just what it says. Structured data raises the odds your facts get parsed correctly and pulled into an answer, because it removes ambiguity. A price, a rating, a founding date, a feature list expressed in schema is far easier for a model to lift verbatim than the same fact buried in a paragraph.

5. Write quotable FAQ and comparison content. Models answer questions, so publish pages shaped like the questions buyers ask: "best X for Y," "X vs Z," "how much does X cost." Lead each answer with a tight, self-contained sentence the model can quote without editing. This works because retrieval engines look for a passage that directly answers the query — give them one they can drop straight into the response with your name attached.

6. Earn third-party citations and reviews. Perplexity and ChatGPT retrieve from the live web, and they trust corroboration. When your product shows up in a G2 category, a Reddit thread, an independent listicle, and a review site, the model sees consensus and repeats it. You cannot self-declare your way into an answer; the engines weight what other sites say about you. Pitch roundups, stay active where your buyers discuss tools, and make it easy for reviewers to cover you. This is often the difference between being known and being recommended by AI.

7. Keep content fresh. Retrieval engines prefer recent pages, and stale facts get quietly dropped. Update pricing, dates, comparisons, and stats on a schedule, and stamp pages with a visible "last updated." Freshness signals that your information is safe to cite today, which is exactly the judgment a model makes before quoting you.

Do all seven and you are readable, quotable, corroborated, and current — the four properties every engine rewards. Skip a few and you leave the answer to a competitor. For the strategic frame behind this, compare Generative Engine Optimization vs SEO.

How to measure your AI visibility

You cannot improve what you cannot see, and AI answers are invisible from your analytics dashboard. There is no "AI referral" line in Google Analytics telling you ChatGPT recommended a rival this morning. You have to ask the engines directly, at scale, and score the results.

That is exactly what a Spike scan does. It scans your visibility free by asking the real buyer questions in your category across ChatGPT, Gemini, Perplexity, Claude, Grok, and Copilot, then scores you 0 to 100 on three things that matter: mention rate (how often you come up at all), rank when mentioned (are you named first or buried sixth), and sentiment (do the models describe you well or lukewarm). It also shows you who AI recommends instead of you — the competitors filling the answer you should own — and hands you a Fix Pack that ships all seven fixes above. You can see a sample report before you run your own.

The score is the point. It turns a vague fear ("are we even showing up?") into a number you can move, and it tells you which of the seven levers to pull first.

A 30-day plan

• Days 1 to 3 — Baseline. Run a scan, record your 0 to 100 score, and note who the engines recommend instead of you. Screenshot the answers so you can prove progress later.
• Days 4 to 10 — Fix the plumbing. Confirm crawlers see real HTML, unblock the AI bots in robots.txt, publish llms.txt, and add schema to your top pages. This is the fastest-moving work and it unlocks everything else.
• Days 11 to 20 — Ship quotable content. Write or rewrite three pages as tight FAQ and comparison answers targeting the exact questions from your baseline scan. Lead each with a one-sentence, liftable answer.
• Days 21 to 30 — Build corroboration. Get listed or refreshed on two review sites, pitch one industry roundup, and seed one honest community thread. Then rescan and compare the score.

Thirty days is enough to move plumbing and content, which retrieval-first engines reward almost immediately. Reputation on memory-blended engines like Gemini keeps compounding after that, so treat the rescan as the start of a monthly rhythm, not a finish line.

FAQ

How do I get my product recommended by ChatGPT?

Be readable, quotable, and corroborated. Make sure ChatGPT's crawler sees real HTML, publish tight FAQ and comparison pages it can quote, and earn mentions on the third-party sites it retrieves from, like G2, Reddit, and listicles. ChatGPT with search assembles answers from the live web, so the more your name shows up across trusted sources, the more often it names you.

Can I pay to appear in AI answers?

No. AI recommendations are earned, not bought. There is no ad slot inside an organic ChatGPT, Perplexity, or Gemini answer — the model surfaces what it retrieves and trusts. Anyone selling "guaranteed placement" in AI answers is selling something the engines do not offer. You move the needle by fixing your site and earning citations, not by paying for a spot.

How long does it take?

Technical fixes — crawlability, robots.txt, llms.txt, schema — can be picked up within days to a few weeks on retrieval-first engines like Perplexity and ChatGPT. Content and third-party reputation compound over one to three months, and memory-blended engines like Gemini reflect your reputation more slowly. Expect early movement in a month and meaningful gains over a quarter.

How do I track it?

Ask the engines the way your buyers do, on a schedule, and score the answers for mention rate, rank, and sentiment. A free Spike scan automates that across six assistants and gives you a single 0 to 100 number to watch, plus the list of competitors winning the answers you want.

Stop guessing whether AI recommends you. Scan your visibility free, get your score across ChatGPT, Perplexity, Gemini, Claude, Grok, and Copilot, and ship the Fix Pack that gets you cited.