Technical SEOSchemallms.txt

Technical & schema.
Make your site readable to machines.

AI engines can't cite what they can't read. This is the plumbing: who's allowed in, what they see when they arrive, and the structured labels that tell them exactly what you are.

Get it looked at →See what an audit covers
The short answerFor humans in a hurry, and for the robots.

Technical SEO for AI search makes sure AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Bingbot and Googlebot) can reach and render your pages. Schema markup (JSON-LD) then labels what's on them: who you are, what you sell, who wrote what, and which questions you answer.

Together they're the cheapest, fastest part of AEO (answer engine optimisation). I write the spec and the JSON-LD, your developer ships it, and it's usually done in two to three weeks.

What I check and fix

The plumbing, one pipe at a time.

AI crawler access

robots.txt, Cloudflare bot rules and WAF settings. A surprising number of B2B sites block GPTBot or PerplexityBot by default and nobody knows.

Rendering

If your content only appears after JavaScript runs, many AI crawlers see a blank page. I check what the bots actually get, not what your browser shows.

Schema that's true

Organization, Person, Service, Article, BreadcrumbList, FAQPage and Product or SoftwareApplication where it fits. Hand-written, validated, matching what's on the page.

Entity consistency

Same name, description and links on your site, LinkedIn, G2, Crunchbase and directories. Mixed signals make models less sure who you are.

Indexing hygiene

Canonicals, sitemaps, noindex accidents, redirect chains, plus Bing IndexNow. ChatGPT's web search has leaned on Bing's index, so Bing matters more than people think.

llms.txt

A plain-text summary of your site for AI tools. It's optional and Google ignores it, but it takes an hour and costs nothing, so I do it.

The honest bit

Does schema help AI search? It's a label, not a magic spell.

There's a lot of nonsense sold around schema. Here's where I stand.

What schema does

  • Removes ambiguity about who you are and what each page is about.
  • Connects your entity to your people, your services and your profiles elsewhere.
  • Makes Q&A easy to lift for engines that quote passages.
  • Is checkable. You can validate everything I ship in Google's Rich Results Test and the Schema.org validator.

What it doesn't do

  • Rescue thin content. Schema on a weak page is a nice label on an empty box.
  • Get you FAQ rich results. Google stopped showing those for most sites in 2023. I still use FAQ markup because it labels Q&A clearly for machines, but I won't sell it to you as a rich-result trick.
  • Work if it lies. Markup that doesn't match the visible page gets ignored, or worse.
  • Replace a good page. It's step two, not step one.
What you get

A spec your dev won't hate.

  1. Deliverable 1

    Crawl and render report

    What each AI crawler can and can't see, with the exact rule or setting causing each problem.

  2. Deliverable 2

    Ready-to-paste JSON-LD

    Schema blocks for each template (home, service, article, about), validated, with notes on where they go.

  3. Deliverable 3

    Fix list, ranked

    Every technical issue ranked by impact and effort. Your developer starts at the top and stops when it stops being worth it.

  4. Deliverable 4

    Check after deploy

    I re-crawl once it's live and confirm it worked. If something broke on deploy, I find it before Google does.

Go deeper

Read the long versions.

The CutWhy schema is what AI reads, not your divsThe CutSchema markup for B2B SaaS: the 6 types that matterThe Cutllms.txt: what it does, what it doesn'tThe CutWhat good technical SEO looks like in 2026The CutInternal linking: the six links every page needsBack to the hubAEO foundations: the full picture
Questions

Technical & schema: straight answers.

If you want to be cited by AI tools, yes. Blocking GPTBot, ClaudeBot or PerplexityBot means those tools can't read your current pages, so they either skip you or rely on old or third-party information. The exception is gated or paid content you don't want reproduced. Block that specifically, not the whole site.
Organization and Person (who you are and who's behind it), Service or SoftwareApplication (what you sell), Article (who wrote what, and when), BreadcrumbList (how pages relate) and FAQPage for genuine question-and-answer content. Most B2B sites need five or six types, not fifty.
Maybe a little, and it's cheap. llms.txt is a plain-text file that summarises your site for AI tools. Google has said it doesn't use it, and no major AI company has confirmed relying on it. It takes about an hour, so I include it, but I won't pretend it's the thing that gets you cited.
By default I write the spec and the code, and your developer ships it. That's usually faster and keeps your dev in control of your site. If you'd rather I ship it directly, that's the hands-on add-on at NZ$500 a month, and works best on platforms like Webflow, WordPress or Shopify.
You can check it yourself. Paste any page into Google's Rich Results Test or the Schema.org validator and you'll see exactly what's marked up. I also re-crawl after every deploy and include a validation report.
Ready when you are

Make your site
machine-readable.

Usually two to three weeks of work. Usually the cheapest part of AEO. Usually the bit everyone skipped.

Start a conversation → Or start with a $497 audit