← Back to all free guides
A free guide by Mariah Brunner ↗ INSTAGRAM ↗ TIKTOK

The New SEO
Is Here.
Are You Ready?

People ask ChatGPT and Claude before they buy now, and AI recommends brands in a completely different way than Google ever did. Here’s my full AI visibility playbook to actually show up.

Everyone used to Google a brand before they bought it. Now they’re asking ChatGPT, Claude and Perplexity, and those assistants pick which brands to recommend in their own way.

The industry calls this Generative Engine Optimization (GEO). You don’t need the jargon. You need four moves, and each one below comes with a prompt that does most of the work.

1. Audit Where You Stand

You can’t fix what you can’t see. Take the questions your customers ask right before buying and run them through Claude, ChatGPT and Perplexity to see if you show up.

Write down 10 to 50 real buying questions for your category. For a clothing brand that’s things like "best travel shorts for men" or "Vuori alternatives." If your brand isn’t in the answers, now you know exactly which questions to go win.

AI Visibility Audit
You are my AI visibility analyst. Goal: find the buying questions where AI assistants should be recommending my brand and aren't. MY BRAND What I sell and my category: [describe] Who buys it: [describe your customer] Main competitors: [list 2 to 4] My website: [URL] STEP 1. Write 25 real buying-intent questions a customer would ask an AI assistant right before purchasing in my category. Mix these types: - "best [product] for [use case or person]" - "[competitor] alternatives" - "[my brand] vs [competitor]" - "is [my brand] worth it" and "[my brand] reviews" - problem questions, like "how do I fix [problem my product solves]" STEP 2. If you can search the web, answer each question the way you would for a real customer today and note which brands you'd recommend and which sources you'd lean on. If you can't search, say so, and tell me which questions I should run myself. STEP 3. Give me a table with: question, brands that show up, does my brand show up (yes / no / unsure), the sources being cited, and your best guess why I'm missing. STEP 4. Rank the 10 questions most worth winning first, based on how close to purchase they are and how beatable the current answer looks. RULES Don't flatter my brand. If you don't know whether I'd show up, write "unsure" instead of guessing. Keep each table cell short.

Do This First

Run your top questions yourself in all three assistants. They lean on different sources, so the questions where you’re missing everywhere are your biggest gaps.

2. Build Pages AI Can Quote

AI assistants tend to quote pieces of a page, so every section has to stand on its own.

  1. Headers are questions. Use the exact words people ask, like "What are the best travel shorts for men?"
  2. Answer first. Put the direct answer in the first 50 words or so under each header, then explain. (Foundation’s GEO guide)
  3. Use tables. Turn pricing and comparisons into tables so the facts are easy to pull out accurately.
  4. Make comparison pages. "Your Brand vs. Your Biggest Competitor" matches how people actually ask AI, and most brands don’t have one.
  5. Add FAQ blocks. Each question a header, each answer short and direct.
Answer-First Page Rewrite
You are an editor who rewrites web pages so AI assistants can find, understand and quote them accurately. INPUTS The question this page should answer: [e.g. "What are the best travel shorts for men?"] My current page: [paste the page text or attach it] Real facts about my product: [features, materials, sizes, price, shipping, guarantees, who it's for] Main competitor to compare against: [name + what I know about them, with sources if I have them] DO THIS 1. Write an H1 that is the exact question a customer would ask. 2. Under it, write a direct answer in the first 50 words that names my product and says who it's best for. 3. Break the rest into sections where each H2 is a real customer question and the first sentence under it answers that question. 4. Build a comparison table (my product vs the competitor) using only the criteria a buyer cares about and only facts I gave you. 5. Add an FAQ block: 5 questions as headers, each with a 1 to 3 sentence answer. 6. List anything you'd want to claim but couldn't, because I didn't give you the fact. RULES Every section must make sense if it's quoted on its own. Use plain words, no hype. Never invent a feature, stat, review or competitor detail. Mark anything about the competitor that I should double-check.
3. Get Other People Talking

This is the most overlooked part. AI cares a lot more about what other people say about you than what you say about yourself. The sharpest GEO folks describe it as consensus: the same thing about your brand showing up across lots of independent places.

Community sites matter most here. Industry citation studies keep finding Reddit among the most-cited sites in AI answers. (Foundation)

So you need to exist off your own site, in Reddit threads in your space, on review sites, in roundups, on podcasts and in YouTube videos. One genuinely good third-party mention can do more than ten of your own blog posts.

Third-Party Mention Plan
You are my digital PR strategist. Goal: get my brand mentioned in the third-party places AI assistants pull from, the honest way. MY BRAND Brand and category: [describe] What makes us genuinely different: [2 or 3 real points] The buying questions I want to win: [paste 3 to 5 from my visibility audit] Time I can spend each week: [e.g. 2 hours] GIVE ME 1. The kinds of third-party sources most likely to shape AI answers in my category: specific subreddits, review sites, "best of" roundups, YouTube channels, podcasts and forums. Label each one as checked (you searched and found it) or a guess. 2. For each, the exact search I'd run to find the threads, posts or roundups where my brand should be part of the conversation. 3. A genuine way to earn a mention in each: what I'd contribute, test, offer or pitch. No fake reviews, no posing as a customer, nothing against a community's rules. 4. A short pitch email I can adapt for roundup writers and podcast hosts. 5. A weekly routine that fits my time budget, with what to do each day. RULES Be specific to my category, not generic marketing advice. If a tactic breaks a platform's rules (like undisclosed self-promotion on Reddit), leave it out.

The Mindset Shift

Stop asking how to rank your page. Start asking how to become the brand that keeps coming up across the whole internet.

4. The Technical Layer

01

Let the right crawlers in

Check that your robots.txt isn’t blocking the bots that power AI search. ChatGPT search uses OAI-SearchBot, Claude uses Claude-SearchBot and Claude-User, and Perplexity uses PerplexityBot. GPTBot and ClaudeBot collect data for model training, which is a separate choice. (OpenAI, Anthropic, Perplexity)

02

Add schema markup

Structured data (JSON-LD) on your product, FAQ and review pages spells out your facts in a format machines read cleanly. Most store platforms add some automatically, so check what you already have.

03

Add an llms.txt file

A short file at your domain root that summarizes your brand and lists the pages that matter, like a site map written for AI. It’s a proposed standard, and Google has said it doesn’t support it. It takes ten minutes, so I treat it as cheap insurance.

Technical Starter Kit
You are a technical SEO specialist setting up the AI visibility basics on my site. INPUTS My brand and what it does: [describe in 2 sentences] My platform: [Shopify / WordPress / Squarespace / Webflow / custom] My most important pages: [list each URL with one line on what it covers] My current robots.txt: [paste it, or open yoursite.com/robots.txt and copy what's there] Product details for schema: [name, price, currency, availability, rating and review count only if real] GIVE ME 1. A robots.txt review. Tell me if anything blocks these crawlers and give me the exact lines to allow them: OAI-SearchBot (ChatGPT search), Claude-SearchBot and Claude-User (Claude), PerplexityBot (Perplexity). Then explain in one line each what GPTBot and ClaudeBot do (model training) so I can decide on those separately. 2. JSON-LD schema for one product page and one FAQ page, filled in with my real info. Leave out any field I didn't give you a real value for. 3. A draft llms.txt file: a short brand summary plus my key pages, each with a one-line description, in the format from llmstxt.org. 4. Plain-English steps for where each of these goes on my platform, and how to test the schema with Google's Rich Results Test. RULES Never make up ratings, reviews or prices. If my platform controls robots.txt for me, tell me where its setting lives instead of telling me to edit the file. Flag anything that needs a developer.