The technical GEO checklist

1. Crawler access

Start here, because it's binary. Check your robots.txt for these user agents and make sure the ones you want are allowed:

Two traps. First, blanket AI opt-outs added during the 2023–24 scraping backlash are still live on many sites — check before assuming. Second, robots.txt is only read from the domain root. If your site lives on a subpath (a GitHub Pages project site, for instance), your robots.txt there is decorative.

Verify, don't assume

Check server logs for actual visits from these agents. A permissive robots.txt with a firewall or bot-protection rule silently blocking the same crawlers is a common and invisible failure.

2. Rendering and markup

Not every AI crawler executes JavaScript, and those that do may not wait for slow client-side hydration. Content that only exists after a framework boots is content at risk.

• ✓ Serve meaningful content in the initial HTML response — static generation or server-side rendering.
• ✓ Test with JavaScript disabled: what remains is roughly your worst-case machine view.
• ✓ Use real semantic elements — one <h1>, ordered headings, real lists and tables.
• ✓ Put important content in text, not baked into images.

3. Structured data that states your identity

This is where GEO diverges most from classic SEO. You are not adding schema to win a rich snippet; you are removing ambiguity about what entity this site represents. A connected graph beats scattered fragments:

• Organization (or LocalBusiness / ProfessionalService) — name, URL, logo, address, contact, area served.
• Person — the founder or key people, linked via founder and worksFor.
• WebSite and WebPage — tying pages to the publisher.
• Service — what you actually sell, with a provider reference.
• FAQPage — question/answer pairs matching visible content verbatim.
• BreadcrumbList and Article on content pages.

Two rules that matter more than completeness. Use @id to link nodes so the graph is explicit rather than a pile of unrelated objects. And never let schema contradict visible content — mismatched FAQ markup in particular is a known source of manual penalties, and it undermines exactly the trust you're trying to build.

The sameAs signal

Your sameAs array — LinkedIn, GitHub, directory profiles, a Wikipedia entry if you have one — is how you tell an engine which external records refer to you. This is one of the highest-leverage and most-skipped fields.

4. Freshness signals

Recency is weighted heavily. Make it legible:

• ✓ Visible "last updated" dates on content pages.
• ✓ Accurate dateModified in Article schema.
• ✓ A sitemap with honest lastmod values.
• ✓ A real refresh cadence — updating substance, not just bumping dates. Cosmetic date changes are detectable and corrosive to trust.

5. llms.txt — useful, oversold

An llms.txt file at your domain root offers a plain-markdown summary of your site for language models. It's cheap to add, it's a clean way to state your canonical description, and we ship it as standard.

But be clear-eyed: adoption by major engines is not established, and it is not a ranking mechanism. Any agency presenting llms.txt as the core of a GEO offering is selling the easiest deliverable rather than the effective one. Crawlability, entity clarity, extractable content and off-site authority do the actual work.

The 30-minute audit

1. Fetch yourdomain.com/robots.txt — are AI agents allowed, and is it at the true root?
2. View source with JS disabled — is the content there?
3. Run the page through a schema validator — does a connected entity graph appear?
4. Check your canonical, og:image and og:url actually match the live URL.
5. Confirm sitemap.xml resolves and is submitted in Search Console and Bing Webmaster Tools.
6. Search your own brand name and check what the top external sources say about you.

If you'd rather have this scored for you, our free AI visibility checker walks through the same ground in a few minutes.

Frequently asked

Keep reading