No agency publishes its own audit. Here's ours.

Our own AI visibility

techhorizonlabs.com is measured by our own scanner — Are you found by AI?, the same engine clients pay for. It queries the live AI engines, scores what they actually say about us, and penalises the same things we penalise everyone else for. We publish the score here each time it moves — up or down.

Same scanner, same scoring, no thumb on the scale. If the number is bad in a given month, it prints bad.

Agent readiness: the machine-readable surface

Most websites are built for eyes only. This one is also built for the crawlers and agents that increasingly do the reading — every surface below is live right now, on this domain:

/llms.txt

A curated map of the site for LLMs and research agents — who we are, what each page covers, how to cite us. The first thing a well-behaved agent reads.

/llms-full.txt

The full text of every indexed page in one plain-text file, regenerated at every build — an agent can read the entire site in a single request instead of forty.

Markdown twin of every page (.md)

Every HTML page has a Markdown sibling: append .md to the path (this page is /transparency.md), or send an Accept header preferring text/markdown and the server negotiates it for you. Same content, no boilerplate, far cheaper for a model to parse — and every HTML response advertises the alternate in its Link headers.

/robots.txt with Content-Signals

Beyond the crawler allow-list, machine-readable usage preferences: search=yes, ai-input=yes, ai-train=yes. We say yes on purpose — being read by AI is the point.

/.well-known/api-catalog

A standard linkset of the public JSON endpoints, so agents discover what they can call here without scraping for it.

/.well-known/agent-skills/index.json

The agent-skills discovery index: what an agent can do on this site, declared rather than guessed.

Structured-data @graph on every page

One linked graph — Organization (ABN-anchored), founder, WebSite, and every page's node, all connected by @id — so machines can tell us apart from every other “Horizon” outfit and quote prices, offers and facts without guessing.

Performance

The site is static HTML behind a CDN — no framework shell, no client-side hydration, nothing to boot before the content paints. Critical CSS is minified and inlined into the page at build time, and the two typefaces are self-hosted, subset and preloaded. That architecture is why the machine-readability above works: fast, plain pages are legible pages.

We could quote you our own lab numbers here, but self-reported speed scores are exactly the genre of claim this page exists to avoid. Test us live instead, on Google's infrastructure, right now:

Run PageSpeed Insights on techhorizonlabs.com →

The current fix list

Every audit we deliver ends with a fix list. Here's ours — the open items on our own file, stated plainly:

• Entity disambiguationIn progress

  “Horizon Labs” is a crowded name, and AI engines sometimes blur us into other firms that share it. The registry-anchored fix — ABN in the Organization node, founder schema, verified sameAs links — is live; we keep monitoring until the engines stop confusing us with anyone else.

• Off-page citations still thinOpen

  AI visibility in 2026 is mostly won off your own site — mentions, directories, third-party citations. Our on-site surface is strong; our off-page footprint doesn't yet match it. That gap is the biggest single drag on our own scan score, and the off-page sprint to close it is queued.

• Category benchmark pagesNew

  First-party benchmark pages built from our scanner's Australian dataset — how whole categories score on AI visibility — are newly launched and still bedding in. New pages take time to earn citations; we'll watch them the way we'd watch a client's.

• Google Business Profile buildoutPending

  Our local machine-readable surface (the profile AI assistants and maps lean on for “near me” answers) is behind the site itself. Buildout is on the list; it hasn't shipped. We'd flag this in any client audit, so it's flagged here.