Your Next Customer Didn’t Google You. They Asked ChatGPT. Here’s How to Show Up When They Do.

A few weeks ago I searched for something in my own niche (not my own brand, just a category I work in), except I didn’t search. I asked ChatGPT.

It gave me a clean answer, three recommendations, and moved on. I never saw ten blue links. I never scrolled past an ad. I got an answer and closed the tab.

Then it hit me: my own potential customers are doing exactly this, right now, about exactly what I sell. And I had absolutely no idea whether I showed up in that answer or not.

For twenty years, “ranking” meant one thing: show up on Google’s first page. Now there’s a second front door, and it works completely differently. People are calling it a few different names: LLMO (Large Language Model Optimization), AEO (Answer Engine Optimization), GEO (Generative Engine Optimization), but they’re all pointing at the same shift: the goal isn’t to be clicked anymore. It’s to be the thing the AI quotes when it answers someone’s question.

This blog is everything I’ve learned trying to actually show up there, including the exact prompts I now run every week.

Why Ranking #1 on Google Doesn’t Mean Anything Here

This is the part that surprised me most: being #1 on Google and being cited by ChatGPT are two almost completely different games.

Traditional SEO optimizes for a crawler that indexes your whole page and ranks it against backlinks, domain age, and keyword relevance. AI answer engines do something different. They’re not trying to rank ten pages. They’re trying to extract one clean, confident answer from whatever they can find, and then move on.

A few things that consistently separate “found” from “ignored” by AI engines, based on research analyzing thousands of AI citations:

• The answer has to be near the top. A large share of what gets cited comes from roughly the first third of a page. If your best insight is buried in paragraph 12, it’s invisible to a system skimming for a direct answer.
• Confident language beats hedged language. Pages that state things plainly: “X is the better choice for Y” get extracted far more often than pages that hedge with “it depends” and “some say.” The AI needs something quotable, not something wishy-washy.
• Consensus matters more than your own claims. If you’re the only place online saying something about yourself, AI systems treat that skeptically. If review sites, forums, and independent sources are all saying similar things about you, the AI gains confidence and cites you.
• Freshness is weighted differently by different engines. Some AI search tools strongly favor content updated recently; static knowledge-trained models care less about your last update and more about whether your content was authoritative enough to make it into training data in the first place.
• Each platform behaves differently. Perplexity runs a live web search every time and always shows its sources. ChatGPT’s search mode leans heavily on Bing’s index. Google’s AI Overviews generally only pull from pages already ranking well organically. You’re not optimizing for one algorithm anymore. You’re optimizing for several, each with its own taste.

None of this replaces SEO. It sits on top of it. If your site isn’t indexed and findable in the first place, none of the AI engines can cite you either.

Prompt 1: The AI Visibility Audit (Run This First)

Before changing anything, you need to know where you actually stand. This isn’t a prompt for a coding tool. It’s a prompt you run directly inside ChatGPT, Claude, Perplexity, and Gemini, one at a time, using your own product or service:

Prompt 1: AI Visibility Audit

I want you to act as a neutral researcher, not as a salesperson for me or anyone else.

Someone is looking for [your category, e.g., 'a CRM for small agencies' / 'a Next.js SEO consultant' / 'an AI blog writing tool']. Based on what you know:

1. What companies, products, or sources would you mention in your answer?
2. Is [your brand name] one of them? If yes, how are you describing it? If no, why might that be?
3. What sources, websites, or types of content are you basing this answer on?
4. What would a brand in this space need to do to be more likely to be mentioned by you?

Be honest, not encouraging. I want to know the real gap.

Run this exact prompt across all four major platforms with 10–15 real questions your customers would actually ask. Write down, for each one: were you mentioned, who was mentioned instead, and what got cited.

This single audit usually reveals one of two uncomfortable truths: either you’re invisible everywhere, or you’re invisible specifically where your competitors aren’t. Both are useful. The second one tells you exactly where to focus first.

Prompt 2: The Island Test Rewrite

Here’s the core problem with most existing content: it was written for a human reading top to bottom, not for an AI extracting one isolated paragraph. A simple way to check this. Call it the “island test”: imagine to imagine ripping any single paragraph out of your page and dropping it on a desert island with no surrounding context. Does it still make sense?

Most paragraphs fail this test. They start with “This approach...” or “It works by...” with no idea what “this” or “it” refers to once it’s been pulled out of context, exactly what an AI extracting a quotable chunk has to deal with.

Use this prompt on any page you want AI engines to cite:

Prompt 2: The Island Test Rewrite

Here's a page from my website: [paste the content]

Apply the 'island test' to every paragraph: imagine each paragraph is read completely out of context, with no surrounding page. Does it still make full sense on its own?

For every paragraph that fails:
1. Quote the original paragraph.
2. Explain exactly why it fails (vague pronouns, missing subject, assumed context).
3. Rewrite it so it stands alone, states a direct and confident answer, and could be lifted verbatim by an AI search engine and still make complete sense.

Keep my voice and meaning. Don't add fluff. Prioritize clarity and a confident, direct answer over hedging language.

This is the single highest-leverage rewrite you can do to existing content. You’re not writing more. You’re making what you already wrote extractable.

Prompt 3: The FAQ & Structured-Answer Generator

AI engines consistently favor content that’s already shaped like a direct question-and-answer. Instead of hoping your prose happens to answer a question cleanly, build the Q&A explicitly:

Prompt 3: FAQ & Structured Answers

Based on this page: [paste content or URL]

Generate 8–10 FAQ-style questions that real customers would type into ChatGPT or Perplexity related to this topic. Phrase them exactly the way a person would ask, not how a marketer would title a blog section.

For each question:
1. Write a direct, confident, one-paragraph answer (40–60 words) that could stand completely alone.
2. Suggest where on the page this Q&A block should live.
3. Generate the FAQPage schema markup (JSON-LD) for all of them, ready to paste into the page.

This does double duty: the visible FAQ block gives AI engines an obvious, extractable answer, and the schema markup gives search engines an explicit, machine-readable signal of what each question-and-answer pair is about.

Prompt 4: The Consensus Gap Finder

Remember the audit from Prompt 1: the part where you found out who gets mentioned instead of you? This prompt turns that gap into an action list:

Prompt 4: Consensus Gap Finder

I sell [your product/service]. Here's what I learned from auditing AI visibility: [paste your findings, who got mentioned, what sources were cited].

Based on this:
1. What types of external sources (review sites, forums, comparison articles, YouTube, industry publications) are most likely to be feeding these AI answers in my category?
2. Where do I have zero presence right now?
3. Give me a prioritized list: which 5 places should I focus on first to build the kind of independent, third-party presence that builds AI 'consensus' around my brand?
4. For each one, suggest one concrete, realistic first action I could take this month.

The uncomfortable truth this usually surfaces: you can’t GEO your way to visibility using only your own website. AI engines specifically look for agreement across independent sources. A glowing claim on your own homepage means little. The same claim showing up on a review site, in a forum thread, and in a comparison article. That’s what builds the kind of confidence that gets you cited.

Prompt 5: For Developers: Make Cursor Set Up the Technical Groundwork

If your site runs on something like Next.js, there’s a technical layer underneath all of this: making sure AI crawlers can actually reach and parse your content in the first place. This pairs well with the SEO audit prompts from my last post:

Prompt 5: AI Crawler & Schema Setup

Act as a Senior Technical SEO Engineer specializing in AI search visibility (GEO/AEO).

Audit my Next.js site for AI-crawler readiness:

1. Check robots.txt for any rules blocking known AI crawlers (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, and others).
2. Check whether key pages are accessible without requiring JavaScript to render the core content.
3. Check page load speed. Flag anything that would risk a crawler timing out.
4. Generate or update an llms.txt file describing what content exists on my site and how it should be interpreted.
5. Generate FAQPage, Article, and Organization schema (JSON-LD) for my key pages.
6. Show exact file paths and copy-paste-ready code for every fix.

The llms.txt file is a newer convention: a plain-text file, similar in spirit to robots.txt, that tells AI systems what your site contains and how to interpret it. It’s not yet universally honored by every platform, but it costs almost nothing to add and signals that your site is built with AI visibility in mind.

What This Actually Looks Like in Practice

After running the audit prompt across my own niche, here’s what came back, paraphrased:

“I don’t have direct knowledge of [my brand]. For this category, I’d typically mention [two specific competitors], based on frequent mentions in comparison articles and community discussions. To be more likely to be mentioned, a brand would need visible third-party coverage (reviews, comparisons, or community discussion) rather than just its own website.”

That answer told me exactly three things: I wasn’t visible, my competitors were getting cited from external sources I had none of, and the fix wasn’t “write more blog posts on my own site”. It was “get mentioned somewhere that isn’t my own site.”

That’s a completely different roadmap than traditional SEO would have given me.

Three Mistakes to Avoid

1. Treating this as SEO with extra steps. GEO doesn’t replace technical SEO, good content, or backlinks , it sits on top of them. If you’re not indexed and crawlable in the first place, none of this matters. Do the SEO fundamentals first.

2. Optimizing only your own website. This is the single biggest mistake. AI engines are explicitly looking for independent agreement across multiple sources. A site that only exists on its own domain, with no outside mentions, looks unverified , no matter how well-written the page is.

3. Chasing every platform equally. Perplexity, ChatGPT, Gemini, and Google’s AI Overviews each weight things differently: recency, schema, Bing indexing, community content. Run the audit, find out where you’re already closest to visible, and focus there first instead of spreading thin across all four at once.

Run the Audit Tonight

Open ChatGPT, Claude, or Perplexity right now and run Prompt 1 with a question your actual customer would ask. Not your brand name, the question. The thing they’d type when they have the problem you solve and no idea yet that you exist.

Then come back and tell me one thing: were you mentioned, or was it someone else?

That answer is the entire starting point of every GEO strategy that actually works.