The Contractor Website AI-Search Checklist

Four things: AI crawlers allowed to fetch your pages, your main content present in the server-rendered HTML, answers written plainly enough to quote, and business facts that match your listings everywhere else. Everything past that is optimization at the margins.

That is genuinely the whole technical requirement. Most contractor sites that don't get cited fail at step one or step two — a security plugin blocking the bots, or a page where the useful content only appears after JavaScript runs. Both are fixable in an afternoon and both are worth more than any schema tweak.

For a local contractor, yes. Blocking these bots removes you from consideration in the answers your customers are reading. The publisher argument for blocking — protecting content that AI would substitute for — doesn't apply when your product requires a crew and a truck.

Check this before assuming you're fine. Managed WordPress hosts, security plugins, and CDN bot-protection presets frequently block AI user-agents by default, and nobody tells the business owner. Open your robots.txt and read it, then check your CDN's bot management rules separately — a bot can be allowed in robots.txt and still blocked at the edge.

The standard set: LocalBusiness (or your specific trade subtype), Service, FAQPage, BreadcrumbList, and Review or AggregateRating where you genuinely have them. Schema helps machines confirm facts about you and earns rich results in ordinary search. It is not a magic AI-citation switch.

No. Google's May 2026 AI optimization guidance explicitly groups AI-specific schema, content chunking, and AI-specific rewriting among tactics that do not help with visibility in its AI surfaces. Implement standard schema.org correctly and stop there.

This matters because a whole cottage industry sprang up in 2025 and 2026 selling 'AI-optimized structured data' packages. There is no such vocabulary. If a vendor's proposal includes custom schema types no search engine documents, you're paying for markup that no consumer reads.

It's optional and low-priority. Add it if your site can generate it automatically; skip it if maintaining it by hand competes with real work. The evidence for it as a visibility lever is weak, but the cost is near zero and the forward-looking case for agent traffic is credible.

What the evidence actually says

• Google's May 2026 AI optimization guide states llms.txt is not needed for AI Overviews, AI Mode, or any generative Search feature; John Mueller has compared it to the old keywords meta tag.
• SE Ranking analysed roughly 300,000 domains and found no statistically significant correlation between having llms.txt and AI citation frequency. Adoption was about 10%, and only one of the fifty most-cited domains had the file.
• Ahrefs checked 137,000 domains and reported 97% of llms.txt files were never fetched by any crawler.
• On the other side: Anthropic recommends llms.txt in its agent-writing guidance, OpenAI publishes the file for its Agents SDK, and Chrome's Lighthouse 13.3 added an agentic-browsing audit in May 2026 that checks for it.
• Honest synthesis: it is agent-readiness infrastructure, not an AI-visibility strategy. Automate it or skip it — do not hand-maintain it in place of earning reviews.

Question as the heading, complete answer in the next 40-60 words, detail after. Models extract self-contained passages, so an answer that only makes sense with three paragraphs of context is effectively unquotable no matter how good it is.

Structural rules that matter

• One H1 per page, then H2s phrased as the exact questions buyers ask. Not clever headlines — literal questions.
• Answer immediately under the heading. Never open with 'In today's competitive market...'.
• Use real tables for comparative data. Tables are unusually easy to parse and disproportionately likely to be cited.
• Keep facts in text, not in images. A price list inside a JPEG is invisible to the systems you're trying to reach.
• Server-render the main content. If the useful text only appears after client-side JavaScript, assume some crawlers never see it.
• Give every important claim a specific number, date, or named source. Vague copy is unciteable copy.
• Include the location and service in the answer sentence itself, so a quoted passage carries your context with it.

Yes — this is the highest-value unglamorous fix on the list. When your hours, phone number, service area, or business name differ between your site, Google Business Profile, Yelp, and old directory listings, assistants either report the wrong version or hedge and name a competitor instead.

• Pick one canonical version of your name, address, phone, and hours. Write it down. Use it verbatim everywhere.
• Search your phone number and old business name to find listings you forgot existed — old franchise pages, defunct directories, a former partner's site.
• Fix or claim what you find. Where a listing can't be corrected, get it removed.
• List your real service-area cities explicitly on your site, in text, not only on an embedded map.
• Remove services you no longer offer. Being described as doing work you dropped costs you credibility and wastes call-handler time.

Anything that involves building a parallel version of your site for machines, or paying for placement that isn't for sale. These are the tactics with the worst evidence-to-hype ratio in 2026.

• Separate Markdown or bot-only page versions. Google and Bing have both indicated this crosses into cloaking. Real policy risk, no demonstrated benefit.
• Custom 'AI schema' types no search engine documents.
• Paid 'get listed in ChatGPT' packages — no provider sells answer placement.
• Keyword-stuffing AI terminology into your copy.
• Mass-generating hundreds of thin pages. It dilutes the pages that work and adds no citable substance.
• Blocking AI crawlers to 'protect' your content, then wondering why you're never mentioned.

Work top to bottom. The first four items are pass/fail — if any of them is broken, nothing below it can compensate. Budget a day for items 1-8 and treat 9-12 as ongoing.

• 1. robots.txt allows GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, and Bingbot.
• 2. CDN or firewall bot rules don't block those same agents at the edge — check separately from robots.txt.
• 3. Main content is present in server-rendered HTML with JavaScript disabled.
• 4. Every important page returns 200, is indexable, and has a self-referencing canonical.
• 5. LocalBusiness (or trade subtype) schema on homepage, contact, and location pages, with correct NAP and service area.
• 6. Service schema on service pages; FAQPage schema wherever a real FAQ is displayed; BreadcrumbList on deep pages.
• 7. Article schema with author, datePublished, and dateModified on guides and posts — and those dates are honest.
• 8. XML sitemap current and submitted; it includes every page you want cited.
• 9. H2s phrased as literal buyer questions, with a complete 40-60 word answer directly beneath each.
• 10. Prices, timelines, service areas, and certifications exist as text — never only inside images or PDFs.
• 11. Name, address, phone, hours, and service area identical across your site, Google Business Profile, and every directory.
• 12. A monthly prompt test logged across ChatGPT, Gemini, Perplexity, and Google AI Mode, recording whether you're named and what got cited.