How to Get ChatGPT and Perplexity to Recommend Your Business

A founder runs an AI-powered resume tool. She's doing fine on Google — ranking for a handful of decent keywords, pulling a few thousand visitors a month. Then she checks her Google Analytics for a month and finds something unexpected: she's getting traffic from ChatGPT, Perplexity, and Gemini. Not a lot — maybe 200 visitors total. But those 200 visitors have a conversion rate that's almost five times higher than her organic Google traffic.

She starts paying attention to why. She asks ChatGPT, "What's the best AI resume tool for recent college graduates?" Her product comes up — cited from a comparison page on her own site that she'd barely thought about.

She builds four more comparison and alternative pages. Three months later her AI-referred traffic is 800 visitors a month, and those visitors account for a disproportionate share of her paying customers.

This is what GEO — "Generative Engine Optimization" — actually looks like in practice. Not some exotic new discipline. Just SEO, applied with awareness of how AI tools actually work.

---

The most important thing to understand about AI recommendations

Here's the misconception that wastes more time than anything else in the GEO space: people believe AI recommends businesses based on what it "learned" during training. They think their brand needs to be "in the AI's memory."

That's not how it works. Not at all.

When you ask a modern AI assistant about which brand to buy, which service to hire, or which tool to use — it runs a live web search. Right now, at the moment you asked. It reads the top results. It quotes what it finds.

Your site is almost certainly not in any AI's training data. It doesn't need to be. The AI finds you (or doesn't) at the moment of the question, using the same web index that powers Google search.

Google confirmed this explicitly in its own documentation on AI features: "Generative AI features are rooted in core search ranking and quality systems." ChatGPT searches the web through Bing. Perplexity searches the web through a combination of sources, many of which pull from Google's index. The founder who shut down her GEO startup because she realized "there's no such thing as a GEO strategy separate from SEO" had it exactly right. The AI is a layer on top of search, not a separate universe.

This should be liberating: you don't need to figure out how to "teach" an AI about your brand. You need to rank in search. Good SEO is GEO.

---

Why AI traffic is worth pursuing despite the smaller volume

A fair question: AI traffic is a tiny fraction of overall search traffic — studies show it at under 1% of total internet attention share. Why optimize for it at all?

Two reasons.

Conversion rate. Data from LendingTree showed AI-referred traffic converting at 4–5× the rate of standard organic search. Fewer visitors arrive, but they arrive pre-qualified by the AI's recommendation. The AI said "you should use this" — so the visitor is already 80% sold before they land.

Audience overlap. Around 95% of ChatGPT users also use Google. Optimizing for AI visibility doesn't require a separate strategy — you're largely serving the same audience through a different entry point. Good SEO earns both.

---

The technique almost nobody uses: see the actual searches the AI runs

Here's something crucial that most GEO advice misses: AI doesn't search for your prompt verbatim. It rewrites your question into one or more different searches — a process called query fan-out. If you ask "has the Empire State Building changed its observation deck recently?" ChatGPT might actually search "Empire State Building observation deck changes 2025 ticketing."

The difference matters. Optimizing for your question's wording and optimizing for the AI's actual search are different things.

You can see exactly what searches an AI runs, using browser developer tools:

For ChatGPT:

1. Run your target prompt in ChatGPT
2. Copy the conversation ID from the URL (the string after /c/)
3. Right-click the page → Inspect → Network tab
4. Paste the ID into the Network search filter → Refresh the page
5. Find the matching request in the list → click it → Response tab → search for queries
6. You'll see the literal web searches ChatGPT executed — often very different from what it says it searched

For Perplexity: Much simpler — click "steps completed" and Perplexity shows its searches openly.

Once you know the exact queries the AI runs when someone asks about your category, the optimization is obvious: use that exact language in your page title, URL slug, H1, first sentence, and H2 sections. This is semantic similarity — the closer your content matches the AI's actual search, the higher the probability you're cited.

Run twenty different prompts about your industry, map all the language the AI uses, and you'll have a complete picture of how to write pages that get pulled into AI answers.

---

What the research actually shows matters (vs. what's hype)

An analysis of over 1.4 million AI prompts identified what actually influences whether a page gets cited. Here's what the data shows:

High impact:

• Semantic relevance to the AI's real search — this is the #1 factor, by a significant margin. Pages that closely match the AI's search query get cited. Pages that don't, don't.
• Conversion-focused pages — landing pages, product pages, and comparison pages get cited at roughly 29% of the time in AI recommendations. Blog posts get cited far less. This is because these page types match the intent of "recommend something to buy" questions, which is the majority of commercial AI queries.
• Answer-first structure — pages that put the key claim at the top, with a short summary, are easier for AI to extract a quote from. AI tools are looking for the cleanest possible answer to cite.

Low impact (the myths):

Google's own documentation on AI features explicitly calls these out as things that do not drive AI visibility:

• Schema markup / structured data — a study of nearly 2,000 pages that added schema saw barely any change in AI citation rates
• llms.txt files — none of the major SEO authority sites (Ahrefs, Moz, HubSpot, Semrush) use these, and it makes essentially no difference
• "Chunking" content for AI — Google's own documentation says there's "no requirement" for this
• Rewriting content specifically for AI — if your content is good and answers the query clearly, AI will cite it without special treatment

Don't spend hours on schema and llms.txt. Match the AI's language. Answer clearly. That's it.

---

Finding AI-style searches inside Google Search Console

You don't need to run searches in ChatGPT to find the queries AI cares about. They're already in your Search Console data, because AI tools search through Google's index.

7-word searches (likely AI-assisted):

^(?:\S+\s+){6,}\S+$

Queries this long are typically written by someone using an AI tool to help them search, or describing a complex situation to get a precise answer. These are the searches AI models generate when doing their query fan-out.

"I am" searches: Filter queries containing the phrase "I am" — these are personal prompts where someone describes themselves to get an AI recommendation ("I am a freelance designer looking for a client management tool that integrates with Figma"). These reveal exactly how your buyers are presenting themselves to AI tools. Mirror that language in your pages.

---

Measuring your AI visibility

Three places to check your AI footprint:

Google Analytics "Ask Analytics Advisor" (beta): ask in plain English, "How much traffic did I get this month from ChatGPT? From Perplexity? From Claude?" If your GA4 is connected, it knows.

GA4's ai-assistant medium: Google Analytics now auto-tags visits referred by ChatGPT, Gemini, and Claude under a medium called ai-assistant. Filter your traffic report for this medium to see the total volume and which pages are earning citations.

Search Console's Generative AI performance report: A newer report showing your impressions inside Google's AI features (AI Overviews) over time, broken down by page, device, and country. Rolling out slowly — check if you have it under Performance.

---

The conversion rate insight that changes the math

Think about what it means that AI-referred visitors convert at 4–5× the rate of standard organic traffic.

If you're currently getting 10,000 organic visitors a month at a 2% conversion rate, you have 200 conversions. If you got 500 AI-referred visitors at a 9% conversion rate, that's 45 conversions. Ten percent of the traffic producing 22% of the output.

As AI search grows, this ratio becomes increasingly significant. Building the asset base that earns AI citations isn't a side project. It's arguably the highest-ROI SEO investment available right now, before the space gets as competitive as traditional organic search.

The asset base: clear, answer-first conversion pages targeting the exact language AI uses when it searches your category. Start there.

---

One often-missed citation factor: link out and cite your sources

A controlled test used a made-up keyword ("Volandasysik") to set up 10 identical pages — same content, same structure — except five linked out to external sources and five didn't. The five that linked out consistently ranked above the five that didn't. The same pattern holds for AI citations: pages that demonstrate research by citing sources are treated as more authoritative.

Add outbound links to credible sources in your content. Cite data. Reference studies. This signals to both Google and AI tools that your content is grounded in real information, not just internal claims.

---

Bing: the overlooked ChatGPT lever

Most AI search optimization focuses on Google. But ChatGPT specifically pulls results from Bing, not Google. This makes Bing Webmaster Tools a direct lever for ChatGPT visibility that almost no one uses.

Connect your site to Bing Webmaster Tools and submit your sitemaps there. Ranking on Bing is significantly easier than Google — less competition, faster indexing. The same five keyword placements work for both. A page that would take months to rank on Google can rank on Bing in weeks — and every Bing ranking is a potential ChatGPT citation.

---

Social media posts as instant AI citation sources

Because AI tools cite high-authority platforms (LinkedIn, YouTube, Instagram, X, TikTok), a well-written social post can end up cited in AI Overviews and AI assistant answers within hours.

One documented example: a person posted on LinkedIn claiming a specific expertise. The same day, Google's AI Overview cited that LinkedIn post as evidence of the claim — citing the LinkedIn post as a source, without any verification. The AI read what was in a high-authority source and repeated it.

The practical play: post about your brand on high-authority social platforms with keyword-first language. A LinkedIn post that begins with your target keyword gives AI tools a citeable source they can find quickly. YouTube video titles and descriptions are even stronger — YouTube is currently the #1 social citation source in AI, having recently overtaken Reddit.

---

The checklist

• Stop assuming AI knows your brand from training — it searches live, every time someone asks
• Use the ChatGPT network-tab technique to see the actual queries it runs for your category
• Check Perplexity's "steps completed" to see searches openly
• Connect to Bing Webmaster Tools — ChatGPT uses Bing, and Bing is far less competitive than Google
• Prioritize semantic relevance — use the AI's exact search language in your title, H1, first sentence, and H2s
• Build conversion pages, not blog posts — they get cited at 29% vs. blog posts' much lower rate
• Cite your sources in content — pages with outbound links to credible sources rank higher in AI citations
• Post keyword-first on LinkedIn and YouTube — social posts on high-authority platforms get cited in AI Overviews quickly
• Skip the schema and llms.txt rabbit hole — the data doesn't support it
• Set up GA4's ai-assistant medium filter to start measuring AI traffic today

---

Next: The Three Page Types That Convert Buyers AND Get Cited by AI →