05 / 07 · SEO and AEO
Found by search engines and cited by answer engines.
The search agent does three things most SEO retainers do not: it diagnoses what is actually blocking indexing before writing anything, it builds programmatic pages where you have a real list of things people search for, and it writes for answer engines as well as search results, so your pages are the ones ChatGPT and Perplexity cite. It publishes on a schedule and reads the results back weeks later.
What it reads
What it analyses
- Why pages are not indexed: thin or boilerplate content, internal links declared but never rendered, stale modified dates, and other things a re-request will never fix.
- Which queries you already rank for but nobody clicks, and which you almost rank for.
- What structure the pages that get cited by answer engines share, and where yours differ.
- Whether the content the agent shipped last cycle has settled enough to be judged.
How it decides
Diagnosis before production. The agent will not publish a page onto a site with an indexing problem it has not explained. Content decisions follow a written tree: does this news or query deserve a full page, an article, a social post, or nothing. Every page ships with a hypothesis about the query it should win and a review date, and pages are judged on impressions, clicks and enquiries, never on having been published.
Example experiments
The shape of a test, written before it ships.
Illustrative, not a promise. Every experiment on your account gets this record before anything changes: the hypothesis, what the verdict is read from, what stays fixed, how long until a verdict, and where a person decides.
Hypothesis 01
Pages that have been unindexed for months share a structural cause that a diagnostic can name, and fixing it indexes them where repeated re-requests did not.
- Measured on
- Indexed state of the affected pages in Search Console after the fix, against the months of re-requests before it.
- Held constant
- Page content itself, so the structural change is what is being judged.
- Verdict after
- Four to eight weeks, indexing is slow.
- Human gate
- Structural changes to the site reviewed before they ship.
Hypothesis 02
Answer-first pages, with a direct definition in the opening line and a short summary block, earn answer-engine citations that the existing long-form pages do not.
- Measured on
- Referrals from AI assistants and citations observed for the target queries, against the matched control pages.
- Held constant
- Topic set and publishing cadence.
- Verdict after
- Two to three months.
- Human gate
- You approve the topic list; the agent may publish on its own once trusted to.
Hypothesis 03
A programmatic set of pages built from a real list of things your customers search for earns long-tail traffic that no single article can.
- Measured on
- Impressions, clicks and enquiries attributed to the set, read monthly.
- Held constant
- The rest of the site's publishing schedule.
- Verdict after
- Three months to a first read, six for a verdict.
- Human gate
- Template and first ten pages approved before the set is generated.
Where people decide
Topic and template approval before anything is generated.
Structural changes to the site ship as pull requests you can review.
Publishing autonomy is earned, page type by page type.
What you get
- 01An indexing diagnosis that names causes, not a list of URLs to re-request.
- 02A publishing cadence with every page carrying a target query and a review date.
- 03Pages written to be cited as well as ranked.
- 04A monthly read on impressions, clicks, citations and the enquiries behind them.