Skip to content

Private preview

Posture

A grade for how machines see your site

Posture reads the surface AI systems consume: discovery, extractability, agent interfaces, rights signals. It grades that surface dimension by dimension and shows the evidence behind every score. The index and the rubric are public.

Contact us to request accessPosture is currently in private preview.

In short

When an AI assistant answers a question about your company, it reads your website first. It checks your robots rules, your structured data, and whatever text it can pull from your pages. Most companies have never looked at what it finds.

That surface has many parts: robots rules, sitemaps, feeds, structured data, agent endpoints. Each can be right or wrong on its own, and they often disagree with each other.

Posture checks all of them, grades the result against a published rubric, and lists the fixes in order of effect. Every finding comes with its evidence.

What Posture does

Grade

One posture score per domain, built from measurable dimensions — weighted by what the evidence says moves AI answers, and open about what doesn't.

Map

A versioned archive of the machine-facing web: who publishes which surfaces, what changed yesterday, who references whom. Your competitive set, diffed over time.

Manage

Evidence-ranked fixes, change alerts on the domains you watch, and an advisory tier where we hold the pen and keep every surface current as the standards move.

What gets graded

Each dimension is named for what a machine is trying to do.

Whether AI systems are allowed to fetch your site, and whether they can. A few lines in a robots file decide which of them ever see it.
Whether your content survives being read by a machine: structure, markup that carries meaning, and a page small enough to fit a fetch budget.
Which machine-facing files you publish: sitemaps, feeds, structured data, agent files such as llms.txt.
What an agent can call directly: documented endpoints, MCP servers, and the interfaces you have chosen to expose.
Whether an agent can complete a task on your site. Unlabelled forms and broken accessibility trees stop an agent the same way they stop a screen reader.
What you permit, stated where machines look for it, with indexing and training treated as separate decisions.
Whether you are identifiable as one company across every surface, so a model resolves you to one company.

Graded on evidence

Much of what this industry sells has never been measured. Some tactics have a measured effect on whether AI systems cite you. Others have none, and are sold anyway.

Posture weights each dimension by the evidence and shows that evidence in the report. Where a popular tactic has no measured effect, the report says so, including tactics we could sell you.

The rubric and the index are public so anyone can check a grade.

Getting access

Posture is in private preview. If you want to see your domain’s grade, or you are watching a competitive set you would like mapped, get in touch and tell us the domain.