AI Optimisation
AI optimisation, or AIO, is the infrastructure work that makes a site readable by machines: crawler access, server-side rendering, structured data and speed. It is the first of the three AI visibility disciplines and there is no point attempting the other two without it.
Can a machine read every word on the page?
AI optimisation asks one question with a mechanical answer: when an AI crawler requests your page, what does it receive? Not what a browser renders after JavaScript executes — what arrives in the HTML response. For a large number of sites the honest answer is an empty shell.
This is infrastructure work rather than content work. Robots directives and firewall rules decide whether the request is served at all. Server-side rendering decides whether the response contains the content. Structured data describes what is on the page, and response time decides whether the crawler waits.
It comes first because the disciplines above it depend on it. Restructuring passages on a site a crawler cannot read produces nothing measurable, and OMAV will say so before quoting the later work.
What AIO covers
Six workstreams. Each one is verifiable from your own server logs.
Crawler access audit
Robots directives, firewall and CDN rules, and bot-management settings reviewed per user agent — GPTBot, ClaudeBot, PerplexityBot, Google-Extended and others — because access is frequently blocked by a setting nobody chose deliberately.
Rendering assessment per template
Every template checked for what the HTML response actually contains, so that client-only rendering is identified where it exists rather than assumed absent.
Server-side rendering remediation
Rendering changed only for the templates where it alters what a crawler receives, because rendering work on templates that already serve content is spend without effect.
Structured data implementation
Schema describing the organisation, services, articles and FAQs, validated against the visible text — schema that contradicts the page is a liability rather than an asset.
Performance and response time
Time to first byte and Core Web Vitals addressed at the template level, with a measured budget rather than a one-off improvement that decays.
Log-file verification
Evidence from your own server logs that AI crawlers are fetching pages and receiving content, per user agent and per template, with dates.
How an AIO engagement runs
Audit access
Robots, firewall, CDN and bot-management rules reviewed per AI user agent, with anything blocking access listed and owned.
Assess rendering
Template-by-template comparison of the HTML response against the rendered page, identifying where content is missing from the response.
Remediate
Access rules corrected, rendering changed where it matters, structured data implemented and validated against on-page facts.
Verify from logs
Server log analysis showing which agents fetched which templates and what they received, dated as a baseline.
Hand over with a budget
A performance and structure budget enforced at build, so the position does not decay after the engagement ends.
AIO next to the disciplines above it
Why AIO is first and what the other two add.
| Dimension | AIO | AEO | GEO |
|---|---|---|---|
| Unit of work | The template and response | The passage | The entity and the source |
| Question answered | Can a machine read it? | Can a machine quote it? | Can a machine name you? |
| Primary lever | Access, rendering, schema | Structure and clarity | Authority and consistency |
| Verified by | Server logs | Extraction testing | Answer and citation tracking |
| Time to effect | 4 to 10 weeks | Weeks to 2 months | 3 to 9 months |
| Depends on | Nothing above it | AIO | AIO and AEO |
What AIO cannot do
AIO does not make content worth citing. A perfectly crawlable, well-rendered, correctly marked-up page that answers nothing will be read by machines and quoted by none of them. It removes the mechanical obstacles and nothing more, which is precisely why it is cheap relative to its effect.
It also cannot guarantee inclusion. Whether a model retrieves and cites a source is decided by systems nobody outside those companies controls. What AIO guarantees is that failure is not caused by your infrastructure, which is a narrower and more honest claim than most of this market makes.
- Verify crawler access from server logs, not from a tool that guesses.
- Client-only rendering means many crawlers receive an empty shell. Assess it per template.
- Schema that contradicts the visible page is worse than no schema.
- AIO is a precondition for AEO and GEO, not an alternative to them.
Questions buyers ask about this
How do we know whether AI crawlers can reach our site?
From your server logs. They record which user agents requested which URLs and what status was returned. Tools that infer access from robots.txt alone miss the most common cause of blocking, which is a firewall, CDN or bot-management rule that nobody chose deliberately for these agents.
Our site is a JavaScript application. Is that fatal?
No, but it needs assessing rather than assuming. Client-only rendering means the HTML response is largely empty and many crawlers see nothing. We check template by template and scope server-side rendering only where it changes what a crawler receives, which is usually a subset rather than the whole application.
Which crawlers should we allow?
That is a commercial decision rather than a technical one, and it should be made deliberately. Allowing GPTBot, ClaudeBot, PerplexityBot and similar agents makes your content available to be retrieved and cited. Blocking them protects the content from being used in training and answers. We present the trade-off; we do not decide it for you.
Does structured data still matter?
Yes, as description rather than persuasion. Schema tells a machine what the page is about and how its parts relate, which reduces the work of interpretation. What has changed is tolerance for mismatch: schema that claims something the visible page does not support is a liability, so we validate against on-page facts rather than generating it blindly.
How long does AIO take?
Typically four to ten weeks. Access corrections are often same-week. Rendering changes depend on your stack and release process, which is usually what sets the timeline rather than the analysis.
Marked up as FAQPage structured data, matching the visible text exactly.
Still not sure this is the right service?
Answer four questions and we will tell you which one fits — or that none of them do.
Tell us what you are trying to fix.
A short conversation about the objective, the constraints and the timing. If we are not the right fit, we will say so.