GEO · 6 min read

A GEO readiness checklist for your website

The checks we run first when auditing a site for AI search: crawler access, HTML rendering, page structure, structured data, prompt fit and off-site mentions.

Most GEO problems are boring. A firewall rule blocks a crawler. A page loads its copy with JavaScript, so the HTML a bot receives is empty. Two pages claim to be the canonical version of the same URL. None of this is visible when you look at the site in a browser, which is why it goes unnoticed for months.

This is the checklist we use at the start of a GEO readiness audit. You can run most of it yourself in an afternoon. We ran the same checks on our own site while rebuilding it, and found several of these issues there too.

Can AI crawlers get in?

Access comes first. If a crawler can't fetch a page, nothing else on this list matters for that crawler.

  • robots.txt allows the agents you care about. At minimum, check Googlebot, Bingbot, OAI-SearchBot and ChatGPT-User (OpenAI), PerplexityBot and Perplexity-User, ClaudeBot and Claude-SearchBot (Anthropic), and Google-Extended if you're comfortable with Gemini using your content. Every vendor documents its agents, for example OpenAI and Perplexity.
  • Rules are in the right group. robots.txt rules apply to the user-agent group they sit under. A Disallow line placed under the wrong group is a common mistake.
  • Your CDN or firewall isn't blocking AI bots. Check bot-fight modes, "block AI crawlers" toggles and custom WAF rules. These override robots.txt.
  • Important pages aren't behind logins or consent walls.
  • There's an XML sitemap. It lists only canonical, indexable URLs, and robots.txt points to it.
  • Status codes are honest. Missing pages return 404, moved pages return a 301 to the new URL, and nothing important returns a 5xx.

A quick test: request a page with a crawler user agent and check the status code.

curl -s -o /dev/null -w "%{http_code}\n" -A "PerplexityBot" https://www.yoursite.com/

A 403 or a challenge page means something upstream is blocking the bot. User-agent spoofing won't catch IP-based blocks, so also look at your CDN's bot settings directly.

Does the content render without JavaScript?

  • The raw HTML contains your main copy. Many AI crawlers don't execute JavaScript (Vercel's 2024 crawler study), so a single-page app that ships an empty <div id="root"> looks blank to them.
  • Titles, descriptions and canonicals are in the initial HTML, and they aren't duplicated or overwritten by scripts after load.
  • Navigation links are real <a href> links so crawlers can discover your other pages.
  • Unknown URLs return a real 404. Many single-page apps return a 200 status and a blank page for any URL, which creates thousands of "soft 404" pages.

To test, fetch the page without a browser and look for your H1 and a sentence from the body:

curl -s https://www.yoursite.com/ | grep -c "a sentence from your hero copy"

If it returns 0, prerender or server-render that route.

Is each page structured so it can be quoted?

  • One H1 per page that says what the page is about.
  • Headings follow a logical order (H2 for sections, H3 inside them) and describe the section. Phrasing them as the questions buyers ask works well.
  • The answer comes first. The first two or three sentences of a page or section answer the question directly. Supporting detail follows.
  • Visible FAQs on key pages cover the questions you hear on sales calls.
  • Claims are specific and sourced. Link to public sources for any outside data. The original GEO research found that citing sources and adding statistics made pages more visible in AI answers.
  • Images have descriptive alt text, and decorative images have empty alt text.

Does structured data describe what's on the page?

  • Organization markup with your legal name, logo, URL and official social profiles.
  • WebSite markup on the home page.
  • Service or Product markup for what you sell.
  • FAQPage markup only where a matching FAQ is visible on the page.
  • BreadcrumbList on inner pages, and BlogPosting or Article on posts, with dates and an author.
  • Nothing invented. Google asks that structured data match visible content (Google Search Central). Fake phone numbers, made-up ratings or FAQ markup with no visible FAQ are worse than nothing.

Is it clear who you are?

AI systems need to connect your name to a category and a set of facts. Make that easy.

  • Your brand name is spelled the same way everywhere, on site and off.
  • A one-sentence description of what you do appears on your home page, your about or services page and your social profiles, in consistent words.
  • Contact details are real and easy to find.
  • Your social and profile links are listed in your Organization markup's sameAs field.
  • An llms.txt file at the root summarizes the site and links your key pages. It's a proposed standard, and Google says it isn't required for its AI features, so treat it as a low-cost extra.

Do your pages match the prompts buyers use?

  • You have a written list of buyer prompts, grouped by category, problem, comparison and trust questions.
  • Each important prompt maps to a page that answers it. Gaps become your content plan.
  • You have pages for comparisons and alternatives where they're honest and useful.
  • Your blog answers real questions in your category, with each post focused on one question.

Are other sites talking about you?

  • You appear in third-party comparisons and reviews in your category.
  • Customers and creators describe your product on the platforms your buyers use.
  • Those mentions use your correct name and category words, so models can connect them to you.

This part takes the longest, and it's where creator programs help most. See how creator content feeds AI search visibility.

Are you measuring the right things?

  • A baseline of how ChatGPT, Perplexity, Google AI Overviews and Copilot answer your prompt list today.
  • A competitor set to compare share of voice against.
  • A monthly re-run of the same prompts, tracking mention rate, citations and how you're described.
  • Search Console and Bing Webmaster Tools set up with your sitemap submitted. Google counts AI Overviews and AI Mode traffic inside the Performance report's "Web" search type.

What should you fix first?

Work top to bottom. Access and rendering problems cancel out everything below them, and they're usually the quickest to fix. Structure and structured data come next. Prompt coverage and off-site work take longer and keep paying off.

If you'd like us to run this audit for you, it's the first step of our GEO program. You can book a 30-minute call or ask for a free GEO readiness check.

FAQ

Frequently asked questions.

How long does a GEO readiness audit take?

The technical checks in this list can be done in a day or two for most marketing sites. Building the prompt list and the competitor baseline takes longer, because it involves running many prompts across several AI platforms.

Do I need to rebuild my site to pass these checks?

Usually not. Single-page apps can often be prerendered at build time so every route ships finished HTML, without changing the design.

Which issue do you see most often?

Content that only appears after JavaScript runs. The site looks perfect in a browser, and an AI crawler receives an empty page.

Work with Dime

Want help getting your brand talked about and recommended?

Dime runs creator campaigns at scale and GEO programs that get brands found in ChatGPT, Perplexity, Google AI Overviews and Copilot. Tell us what you are launching and we will reply within 24 hours.