AI SEO Services That Fix and Prove Agent Visibility
AI SEO Services Built on Measurement, Not Guesswork
Monitoring tells you that you're invisible. It does not make you visible. Most of what's sold as "AI SEO" or "AEO" today is a dashboard that watches AI answers mention you or not: useful information, and a dead end on its own. We do the other half. We find what an AI agent gets wrong about your business, fix it, and measure the fix against the same test that found the problem.
What "AI SEO" actually means right now
The umbrella terms (AEO, answer engine optimization; GEO, generative engine optimization; AI SEO) all point at the same underlying shift: buyers are asking assistants about businesses before they ever load a website themselves, and those assistants answer from whatever they can fetch and read in a few seconds. There is no ranking algorithm to reverse-engineer here the way there was for classic SEO. There is a much narrower, testable question: when an AI agent reads your homepage, does it get your facts right? We answer that question with an experiment, not an opinion.
What the engagement includes
The package runs in one sequence, every time, and each step produces something you keep:
- Audit. We send real AI models at your live site, cold, with no framing, and record what they get right, what they get wrong, and what they can't find at all. Pricing is the most common gap: on real client sites, an agent could not determine pricing on 48 of 48 recorded visits, and that result reproduced across three independent passes (Notes 12, 13).
- Fact profile. You answer the questions a buyer would actually ask you on the phone: what you sell, buy-or-rent, service area, how long you've operated, what your credentials mean for the buyer, who you serve, what happens after the sale. A named person signs it off before anything generates.
- Eight generated surfaces. A homepage fact block (the mechanism
that moved the number, see below), a where-to-go index for agents, a machine-readable
fact sheet at
llms.txtandllms-full.txt, an identity card, robots.txt pointers, structured data for search, and a hosted, notarized version-history page. All eight are generated from the one signed-off profile, so none of them can contradict each other. That is a real risk, not a theoretical one: three of five audited client sites contradicted themselves across their own surfaces, one of them quoting a price six times higher in its fact sheet than on its own pricing page (Notes 28, 33). - Before-and-after receipt. We re-run the identical audit (same script, same models, same day of week) against the baseline we froze before installation, and hand you both sets of numbers.
- Monitoring. Monthly re-audits catch drift when a page changes and the fact sheet doesn't, or vice versa. When a fact changes, one update regenerates every surface and adds a version to the notarized history.
What we measured, in short
The headline number: with the fact block in place, agents recovered 78% of a site's key facts, against 30% for the same site with no block, same six models, same day. Agents also stopped after an average of 1.1 page fetches instead of hunting through the site (Note 41). We do not round that up: it was one site, six models, two runs, and we publish the caveat alongside the number every time we cite it.
Separately, we found that publishing a machine-readable fact sheet and hoping an
agent finds it is not a substitute for this. Agents reached an unpointed
llms.txt in 0 of 18 test visits (Note 15), and even when they did fetch a
file containing a real price, none of ten runs reported that price in their final
answer (Note 19). The fix that worked was putting the fact on the page itself, not
routing an agent to a file. See the full mechanics on our
llms.txt guide.
Pricing
| Tier | Price | What it covers |
|---|---|---|
| Free — See what AI sees | No charge | One model reads your homepage. You get the recovery score and the list of answers it could not find. No fix package. One audit per domain every 30 days. |
| Audit + Kit | 149 USD once | The full six-model audit, the generated fix package for every surface, and your notarized version page. Bought once, yours to keep, including the source files. |
| Agent-Ready | 149 USD/mo | The kit, plus monthly six-model monitoring, drift alerts, dashboard, stamp history, and regeneration whenever a fact changes. |
| Pro | 299 USD/mo | Everything in Agent-Ready across three sites, plus competitor benchmarking and priority re-audits. |
| Managed | From 500 USD/mo, talk to us | We install it, run the audits, and sit down with you to work through contradictions and sign the facts off. Scoped in conversation; no checkout. |
These are our published starting prices, not a market benchmark. Scope conversations determine the Managed tier, and multi-location or agency white-label arrangements are priced separately.
Who this is for
Businesses whose buyers now ask an assistant before they ask a human: contractors and trades, manufacturers and equipment suppliers, professional services firms, and agencies who want to install this for their own client book under a managed engagement. If an AI agent answering a question about your business would currently get your price wrong, miss it entirely, or confuse you with a different company under a similar name, this is the fix. That kind of near-miss happened to real client names in our own testing, sending a buyer's agent toward a URL that either doesn't exist or belongs to someone else (Note 14).
See it before you buy it
Run a free audit and see your own recovery score first. When you're ready for the fix, the monitoring dashboard and the packages above are both one form away.