How to monitor indexation health with an AI agent
Sitemaps, index coverage and freshness for the pages that earn money, with the drifts caught before they become traffic losses.
IntentKeep the revenue pages indexed and freshly crawled, and catch coverage regressions while they are still cheap.
- 5
- Steps
- 7
- Tools
- 4
- Rules
- 3
- Failure modes
When to use it
Use it when
- Weekly on large or fast-changing sites.
- After deploys touching routing, canonicals or sitemaps.
- When organic-decay-triage points at coverage.
Do not use it when
- Small stable sites: quarterly is plenty, weekly is noise.
- As a crawler substitute: this reads Google's view via Search Console, not the site's own logs.
What it needs first
Required
Preconditions
- Sitemaps submitted and the property ticked.
- The money-page list known, or derivable from clicks.
Inputs
watchlist- URLs that must stay indexed.Default: top pages by organic clicks.
Procedure
In order. Every tool named is one the gateway ships, and links to its reference.
Read sitemap health: submitted, read, errors, last processed.
Track sitemap deltas since last run: what appeared, what vanished.
Inspect the watchlist URLs in bulk: indexed, canonical, last crawl.
Check crawl freshness against each page's change cadence.
On very large sites, sample sections instead of exhausting quota.
Decision rules
The observable condition, what it lets you conclude, and what takes the conclusion back.
When
A money page reports indexed with a Google-selected canonical different from the declared one
Conclude
Canonical conflict: the page competes with itself; highest priority.
When
Sitemap URLs dropped sharply since the last delta
Conclude
Generation problem upstream; quote the counts and the sitemap file.
When
A watchlist page's last crawl exceeds twice its change cadence
Conclude
Staleness finding: Google serves an old version of a page that changed.
When
Coverage errors concentrate on one URL pattern
Conclude
One routing or template cause; name the pattern, not each URL.
Evidence every conclusion must carry
- Every finding quotes the inspection result verbatim: state, canonical, crawl date.
- Deltas carry both counts and the file they came from.
Where the agent stops
- Fixes live in the site's code and infrastructure; the skill delivers the report.
- Requesting indexing at scale is rate-limited by Google and stays a human call.
How it goes wrong quietly
The cases where the analysis is wrong and still looks right. Read them before trusting a number.
- Inspection quotas cap bulk checks; on big sites the read is a sample and must say so.
- Search Console lags real crawling by days; a deploy today judged tomorrow reads as broken.
- Intentional noindex pages inflate error counts; the watchlist must encode intent, not just traffic.
What the answer contains
- 01Watchlist status
- Each money page: indexed, canonical, last crawl.
- 02Regressions
- New coverage problems since last run, by pattern.
- 03Sitemap report
- Health, deltas, processing dates.