audit_content

Cost: 12 credits.

audit_content is the bridge between your Search Console data and your crawl. It takes the queries your site already earns impressions for, finds the page Google shows for each one, and reports the queries whose words appear nowhere in that page's title or h1. It is synchronous and returns its findings immediately. Run both pull_gsc_data and crawl_site first — with either one missing the tool says which to run and charges nothing.

What it does

It joins two things you have already paid for and calls no outside service. From your latest pull it takes each (query, page) row; from your latest crawl it takes each page's title and h1s. A query counts as said when every one of its meaningful words appears somewhere in the title or an h1 of that page, matched case- and accent-insensitively, and anywhere inside a word (so "shoe" is found in "shoes").

A short bilingual list of filler words (the, and, for, ve, ile, bir, and a few more) is ignored — their presence in a title says nothing about whether the page covers the query.

Your own brand words are ignored too. A query whose only missing words are your site's own name is dropped from the list entirely, and a query missing your brand and something else keeps only the something else. The reply says how many were dropped that way. This is not a courtesy: reporting a firm to itself for not repeating its own name in its titles is a finding nobody can act on, and a list of those crowds out the ones you can. Everything else has to be there.

Both the title and the h1s are read, deliberately. A title trimmed to fit the search result routinely drops a qualifier the heading keeps, and reading the title alone would report every such page as a problem — a list of findings you cannot act on is worse than no list. Results are ordered by impressions, biggest missed opportunity first.

How much it could check

The answer always states how many of your query/page pairs it was able to check and across how many crawled pages, because this tool joins two measurements with different reach. A crawl that covered 30 pages of a 300-page site produces a clean report that means very little — and "no mismatches found" would read exactly like a site whose titles are all correct. When queries point at pages your crawl never reached, the reply says how many and points you back at crawl_site.

Example

Ask your MCP client in plain language:

Which pages on example.com rank for things their titles don't mention?

Then hand a finding straight back to your assistant:

Rewrite that page's title so it covers the missing words.

Input

Returns

The coverage line comes first, above the findings, together with the window the Search Console figures cover, when that data was pulled, when the crawl was taken, and how many queries were dropped as brand-only. A ratio that changes how you read the list has to arrive before the list, not under it.

Then one line per mismatching query — the query, the page ranking for it, its impressions and clicks, the words that are missing, how many of the query's words the page does carry, and the page's current title — biggest opportunity first. The list is capped; past the cap the reply says how many more pairs mismatch. If nothing mismatches, it says so (and you are still charged for the delivered analysis).

The reply opens by naming the crawl it judged — a short job id, the date it was taken, how many pages it covered and how many URLs it skipped. If you did not pass job_id, it also says so and tells you how to choose a different crawl. That line exists because the newest crawl is not always the widest one: a narrow re-crawl of one section becomes the newest, and an audit that silently judged it would charge full price for a fraction of your site.

Inherited limits

This audit sees only what its two inputs brought back. A pull stops at 100,000 (query, page) rows per window at the very most, and usually sooner at a storage budget measured on the rows themselves, so on a large property a mismatch outside the top rows Google returned is not visible here — the analysis names the row count the window was cut at when that happens. The window also ends 3 days before today rather than running up to it. On the crawl side, only pages your last crawl actually reached can be checked, which is what the coverage line above is for.

A mismatch is a statement about the page as crawled: a title you fixed after that crawl still shows up until you crawl again, which is why the crawl's own date is printed with the findings.