Function review 1 — SEO / AEO

What follows is that, for the first function. Nothing here is estimated — the check names are read out of the engine's own source, and the ship counts are counted across all 261 workspaces.

The thirty-second answer

Asked what the product does for a ski shop in Breckenridge, the honest reply is:

1. What it does on its own today

Not "can in principle" — what it has actually done, counted across the network:

After each publish it rebuilds the sitemap and pings IndexNow, so the page is findable rather than merely present. A snapshot is taken first, so any publish can be rolled back.

47 further items are marked shipped and are deliberately NOT counted above. Those pages were built by a site generator and back-recorded into the ledger so it matched what was actually live — real pages, but not the engine publishing on its own. Counting them would have made this number 302. The distinction matters most in the page row: most of the excluded items are enquiry forms.

The 9 checks it runs on SEO

• Title tag
• Meta description
• Heading hierarchy
• Canonical tag
• XML sitemap depth
• robots.txt
• Search Console verified
• Title/H1 keyword alignment
• Breadcrumb schema

The 17 checks it runs on AI Search — the AEO half

This is the part most competitors do not have. It covers both how answer engines extract you and how agents read you:

• llms.txt
• AGENTS.md
• Structured data (JSON-LD)
• Rich answer schema
• Agent read API
• ai-plugin manifest
• Markdown negotiation
• Agents allowed in robots
• Open Graph complete
• AI answer identity (sameAs)
• Answerable Q&A (FAQ schema)
• Question-based headings
• Citable answer passages
• Freshness signal
• Bylined authorship (E-E-A-T)
• Cites sources (trustworthiness)
• Identity & authority in schema

A real site, start to finish

Everything above is counted across the network. This is one site, so the quality can be judged rather than described — maxelmagnets.com, onboarded by Tim on 18 August.

Technical scores 11 out of 11 — that is the part the engine can see and fix without help. The failures cluster in AI Search and Content, which is where a human has to supply something the engine cannot invent.

Three of the nine failures, with the fix it wrote

What it then produced for this site

Read that table carefully, because it is the honest answer. The enquiry form is marked shipped, but its own record says a site generator built it and it was back-recorded — and maxelmagnets is not on the publish whitelist, so the engine could not have shipped it. On this site the engine published nothing by itself. It produced four pieces — two content briefs, a lead magnet and a positioning list — and all four are waiting on a person.

That is not the flattering version. It is the version that survives Tim opening the site himself, and it sharpens rather than weakens the case: the engine is demonstrably strong at the technical layer it is allowed to touch (11 of 11), and everything past the page is still a draft looking for a human.

2. What it needs a human for

This is the honest centre of the review. The engine is strong at the technical on-page layer and produces a great deal of everything else that a person must then carry. That is not a defect to hide — it is the line to state plainly to a customer.

3. What we would want automated later

• Publishing to a customer's own site — a small connector or CMS plugin would move most of column two into column one.
• Social posting once accounts are connected — 740 posts are drafted and idle.
• Measuring the effect. We publish and then cannot see whether it worked. Rankings, AI citations and mentions are all unmeasured.
• Entity consistency across sites — 55 of our own sites were recently found calling themselves the wrong name, which is an AEO problem as much as a cosmetic one.

4. Is it commercial?