How to do generative engine optimization: a practical GEO checklist
The short answer: generative engine optimization (GEO) is the work of making your site reachable, readable, and quotable for the AI systems that assemble answers, ChatGPT, Perplexity, Google's AI Overviews and AI Mode among them. The proven part is mechanical: let the AI crawlers in, serve your content without requiring JavaScript, and put a direct answer at the top of each page under a heading that matches the question. The rest, brand mentions and llms.txt included, ranges from well-evidenced correlation to unproven convention, and an honest checklist labels which is which.
What is GEO, in one paragraph?
When someone asks an AI assistant a question, the assistant does not read one page and summarize it. It runs its own searches, retrieves candidate pages, and stitches an answer from the passages that answer best. Ahrefs measured this in a March 2, 2026 study: a single prompt fans out into an average of 9 to 11 hidden sub-queries (Ahrefs, query fan-out). GEO is everything you do so that your pages survive that pipeline: get retrieved, get parsed, get quoted. Our guide on what GEO is and is not covers the vocabulary; this post is the to-do list.
Step 1: confirm the AI crawlers can reach you at all
This is the step with the most failures and the least glamour. Every downstream tactic is worthless if GPTBot, ClaudeBot, PerplexityBot, and Google-Extended get a 403 at your front door. Sites block AI crawlers by accident constantly: a CDN bot-protection toggle, a robots.txt copied from a template, a firewall rule left over from a scraping scare. We measured how common this is across the live web in our AI crawler blocking study.
The check is not "does my robots.txt allow them." Your robots.txt can say yes while your CDN says no. Fetch a page as each bot (curl with the bot's user agent is a fair first approximation) and confirm you get a 200 and your actual content back. OpenAI publishes its crawler user agents and IP ranges at platform.openai.com/docs/bots; Google documents Google-Extended in its crawler documentation. This is exactly what the bot-reachability checks in our scan do: request your pages as the real AI crawlers and report who gets in and who gets a wall.
Step 2: make sure your content exists without JavaScript
Most AI crawlers fetch raw HTML and do not execute scripts. If your page needs JavaScript to render its content, an AI crawler sees an empty shell, and as far as the answer engines are concerned your best material does not exist. View your page source (not the browser inspector, the raw source) and check that your headings and body text are actually in it. If they are not, that is a rendering problem to fix before anything else on this list.
Step 3: write passages a machine can lift
Answers get extracted at passage level, not page level. A strong fact buried mid-paragraph under a vague heading rarely survives the cut into an AI answer. The fix is structural, not clever:
- One question per section, stated as the heading.
- The direct answer in the first one or two sentences under that heading, written to stand alone if quoted.
- Specifics over adjectives: numbers, names, dates, each with a source.
- Supporting detail after the answer, not before it.
Google's own guidance for AI features says the same thing in its own words: there is no special markup or GEO trick, the advice is unique, helpful content that satisfies the query (Google, AI features and your website). Our scan grades this as a citable-facts problem: does each key page contain self-contained, sourced statements an engine could quote?
Step 4: get named in the places machines read
In December 2025, Ahrefs studied 75,000 brands and found that branded web mentions correlate with AI visibility at roughly 0.66 to 0.71 across platforms, while backlinks came in around 0.19 (Ahrefs, brand visibility factors). Being named accurately on review sites, industry lists, local press, and YouTube tracks with showing up in AI answers far more strongly than link building does.
Two honest caveats. This is correlation, and Ahrefs says so. And consistency matters: if the machines find three spellings of your name and two versions of your address, none of them earn trust.
Should you publish an llms.txt?
If it takes ten minutes, yes. If it replaces any step above, no. The format is an informal convention from llmstxt.org with no official specification and no documented support from the major engines. Our own measurements put adoption at 11.9 percent of the reachable Tranco top 1,000 as of July 2026. It has no proven downside and it forces a useful exercise, deciding which pages matter and describing them plainly. Our guide on what an llms.txt file actually does is the honest version. Treat it as a courtesy document, not a ranking lever.
What is proven and what is correlational
| Checklist item | Evidence status | |---|---| | AI crawlers get a 200 and real content | Mechanical. Blocked means invisible, no debate. | | Content renders without JavaScript | Mechanical. Most AI crawlers do not execute scripts. | | Answer-first passages under question headings | Strongly supported by how retrieval and quoting work; Google's guidance points the same way. | | Consistent brand mentions across the web | Correlational (Ahrefs, 75k brands, Dec 2025). Strong signal, not a guarantee. | | llms.txt | Unproven convention. Cheap, harmless, not a substitute for anything above. |
Nobody outside these companies can promise you a citation, and anyone selling guaranteed placement in AI answers is selling something that does not exist. What you control is eligibility.
Where does your site stand today?
Steps 1 and 2 are checkable in an afternoon by hand, or in about a minute by machine. Run a free Visibility Scan at auditlamp.com. It reads your site the way the AI engines do, runs it against our 164 graded checks including crawler reachability and quotable answers, and names every failing check in plain language, worst first.
Sources: Ahrefs, "What is Query Fan-Out? Understanding the Hidden Queries Driving AI Search," March 2, 2026 (ahrefs.com). Ahrefs, "Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)," December 12, 2025 (ahrefs.com). Google, "AI features and your website" (developers.google.com). OpenAI crawler documentation (platform.openai.com). AuditLamp, AI crawler blocking study (auditlamp.com); llms.txt adoption note (auditlamp.com).