How the daily check works

What is read each day

Every live prompt is asked on Google AI Overviews, ChatGPT, Gemini and Perplexity, and every cluster keyword is checked on Google. Every read is made for your market, the US or the UK. Claude is added on alwaysmentioned and above.

The check runs once a day. The top of the Overview says when the last one finished, and Settings shows when the next is due. Reads that did not come back on a day are not filled in later, and a day before your first check cannot be read.

How each engine is read

Google AI Overviews, ChatGPT and Gemini are read from the product a buyer uses. Perplexity is asked through its model directly, not read from its app; so is Claude on the higher plans.

When a Google results page carries no AI Overview for a prompt, we record no answer for that day, not a miss.

What we keep of each answer is what each engine said, with link addresses taken out of the text, and the pages it cited, kept separately. Both are kept for the life of the account.

What counts as naming you

An answer names you when its text uses one of the names in Settings, under Names we match: your brand, and any other spellings we have set. A name that appears only inside a link's web address does not count.

We set those names, not you, because each one changes what being named means. To change them, use Ask us to change these in Settings.

The other brands in each answer are read from the same text and matched the same way. They make up share of voice and Who is named. Being cited is separate: a page of yours an answer cites is counted on Cited pages, whether or not the text named you.

No answer is not a miss

If an engine gives no answer to a prompt on a day, that read is left out of the figures for that day. It does not count against you. So each engine's figures are out of the days it answered, which is why one engine can show fewer days than another for the same prompt.

Complete, partial and failed checks

Each day's check ends in one of these ways:

• Complete: every read came back.
• Partial: some reads failed even after retrying, or the other brands could not be read in some answers.
• Failed: none of the reads came back, or the check stopped before it finished.

A read fails when the request for it errors or runs out of time. A read that fails on a temporary error is retried while the check has time left.

Every dashboard page with dates names, at the top, each check in its range that lost reads - partial or failed - with the day and, where it can, the engine or the Google keyword positions that lost them. A partial check's note ends: "Some reads did not come back; they are left out of the figures, not counted as misses."

That is the rule everywhere: a read that did not come back is left out, never counted as a miss against you. It does mean that day's figures rest on fewer answers.

If only the other brands could not be read, every read came back. Those answers still count towards whether you were named, but are left out of share of voice and Who is named, and those pages say how many answers were left out.

When today's check failed, the Overview says so and says which day your figures run to. A check that failed part way through says it did not finish, and the reads it stored before it stopped are in your figures. On one cluster's page, Latest answers shows the last check with an answer, and says why when a later check in the range failed.

Google positions

Each cluster keyword is checked in Google's organic results for your market, down to position 20. The position is where your site's best-ranking page sits, and the dashboard shows which page that is.

Outside the top 20 is a reading, not a gap: the Keywords CSV leaves the position blank, and an average position counts it as #21, so dropping out makes the average worse rather than better.