GEO Quick Start.
The 7 steps that get your brand cited in AI search — ChatGPT, Perplexity, Claude, and Gemini.
Generative Engine Optimization (GEO) is the practice of optimizing your site so AI search engines cite and recommend your brand when users ask questions in your category. It is not SEO with a new name — it runs on different signals. The seven steps below close the most common gaps.
Do these two checks before anything else. They take under five minutes and tell you more about your AI search visibility than any paid tool.
Ask the question your customer asks.
Open ChatGPT. Ask: "What are the best [your product or service category]?"
If your brand doesn't appear in the answer, you have an AI search visibility gap. That's the diagnosis.
Look for two strings in your HTML.
Right-click any FAQ page → View Page Source → search the raw HTML for:
FAQPage— structured Q&A markupsameAs— your brand's entity profile
Missing either? You've just found your two biggest gaps — before spending a dollar.
Both checks run against the raw HTML response, not the rendered DOM — the same view AI search crawlers see when they retrieve your page.
Many sites are accidentally invisible to AI search before optimization even begins. Training crawlers and search crawlers are different systems. The ones that decide whether you show up in answers are the search bots.
Open yourdomain.com/robots.txt
Confirm there are no Disallow rules for OAI-SearchBot, ChatGPT-User, PerplexityBot, or Claude-SearchBot. Many sites carry over generic AI-blocking rules without realizing they include the search bots too.
Bot Fight Mode is on by default.
If your site runs on Cloudflare, Bot Fight Mode is enabled on every plan — including free — and silently blocks AI crawlers. Go to Security → Bots and add an explicit Allow rule for the user agents above.
OpenAI developer documentation; Cloudflare Bot Fight Mode documentation; Originality.AI 2024 — GPTBot blocked by 35.7% of the top 1,000 sites.
Most sites have FAQ content. Almost none have FAQPage schema. Without it, your FAQ sections are visible to users — and invisible to AI retrieval.
Structured Q&A markup in raw HTML.
FAQPage schema tells AI retrieval systems: "This content is a direct answer to a specific question." It maps each question to its answer in a format the model can extract whole.
View Page Source → search.
FAQPage— schema is presentapplication/ld+json— schema is server-rendered
If neither shows up in the raw HTML, the gap is real.
Schema deployed via GTM or any JavaScript layer is invisible to AI crawlers.
GPTBot, OAI-SearchBot, and PerplexityBot fetch raw HTML and move on — no JavaScript execution. Schema injected client-side simply does not exist as far as the retrieval system is concerned. Verify the markup appears in the initial HTML response, not the rendered DOM.
Relixir, July 2025 — analysis of 50 B2B and ecommerce sites: pages with FAQPage schema achieved a 41% AI citation rate vs 15% without — 2.7× higher. Vercel: 500M GPTBot fetches confirmed no JavaScript execution.
AI models build a brand profile from multiple signals. Inconsistency creates a confused entity — and a confused entity is a low-recommendation entity.
sameAs— URLs for every social profile (LinkedIn, IG, Facebook, X)logo— direct URL to your brand logodescription— concise, matches your About pagename— exact, consistent across schema, About, and on-page mentions
Search the raw HTML for sameAs.
If it's absent or only lists one or two profiles, this is the gap. The model literally has nowhere to anchor your brand identity beyond your domain.
High recognition. Low recommendation.
In the 10,128-brand study, Spotify scored 95/100 on AI brand recognition but only 10/100 on recommendation rate for radio queries — the model classified Spotify as a music platform, not a radio platform, due to inconsistent entity signals. Mangools: 75/100 recognition, 7/100 recommendations. Same pattern.
2025 study of 10,128 brands across 600,000+ ChatGPT responses — recognition vs recommendation gap traced to entity-signal inconsistency.
AI retrieval pulls citations from the top of your page — not the middle. Heading = the model's interpretation of the user's question. The paragraph right under it = the answer.
- Heading — states the topic or question
- Direct answer — 40–60 words, no links, no qualifiers
- Supporting evidence — data, examples, context
Citation winners are ~2× more likely to use definitive language ("is defined as," "refers to") vs hedging.
Hedging language signals uncertainty. The model passes over uncertain sources for confident ones.
Growth Memo — 1.2M ChatGPT responses, 18,012 verified citations; ALM Corp — 2M ChatGPT sessions. Links inside answer capsules reduce citation probability.
Factual density is a confirmed GEO signal. AI retrieval systems treat quantified claims as authoritative — and extract them at higher rates than unsourced equivalents.
"Many brands struggle with AI visibility."
"FAQ schema can improve citation rates."
"A 2025 study testing 10,128 brands across 600,000+ ChatGPT responses found the median recommendation score is single digits."
"A July 2025 study found pages with FAQPage schema had a 41% citation rate vs 15% without."
One sourced, specific data point per section. It doesn't have to be your own research — cite third-party studies clearly. The Princeton GEO study (Aggarwal et al., ACM KDD 2024) found that adding citations and statistics boosted source visibility by up to 40% in generative engine responses.
Aggarwal et al., "GEO: Generative Engine Optimization," ACM KDD 2024 — peer-reviewed analysis of 10,000 queries testing GEO content strategies.
You can't improve what you don't measure. This measurement requires no subscription, no scraper, and no technical access — just discipline.
- Open ChatGPT. Use the same account each time for consistency.
- Ask: "What are the best [your category]? Give me your top recommendations."
- Record whether your brand appears, and in what position.
- Run the same query in Perplexity. Then Claude. Then Gemini.
- Screenshot results. File the screenshot with the date.
- Whether your brand appears at all
- Which competitors appear
- Position in the answer (1st, 2nd, last)
- The exact language used to describe each brand
After implementing each fix, track the same queries weekly and note when your brand first appears — that delta is the most direct proof of GEO progress you can produce.
This is the most direct measure of GEO progress. It requires no tools, no subscription, and no technical access — only consistency.
Every step here is a job — not a fix.
Content drifts. Schema breaks. New pages publish without FAQ markup. Competitor signals change. The 7 steps above aren't one-time work — they're a monitoring problem. That's what MeetGEO is for.
- Independent study: 10,128 brands, 600,000+ ChatGPT responses (2025)
- Relixir — 50-site FAQPage schema study, July 2025
- Growth Memo (Kevin Indig) — 1.2M ChatGPT response analysis, Feb 2026
- ALM Corp — 2M ChatGPT session analysis
- Aggarwal et al., "GEO: Generative Engine Optimization," ACM KDD 2024
- OpenAI developer documentation — OAI-SearchBot / GPTBot distinction
- Vercel — 500M GPTBot fetch analysis (no JS execution)
- Originality.AI — top-1,000-site robots.txt analysis, 2024