The visibility method

The visibility pod answers one question and then acts on it: where does the site stand in Google’s AI answers and in the assistants’ answers, who is named instead, and what changes that. The method is Tandem’s; the suppliers, models and mechanisms behind a measurement are never named, and every figure has a file.

The opening sequence

Six steps, each computed from the one before, before any other work on a project. One step at a time; no later step is described while an earlier one is open, and none is skipped.

1. Read the site the way an assistant’s crawler reads it (Site Auditor): the reading as a document, one task per finding worth fixing, with the pages, the consequence for being quoted, the fix, and whose morning it is.
2. Measure how Google’s AI answers speak about the brand and its category (Answers Analyst): which questions get an answer, who is named, what the answers cite.
3. Agree the question set with you, the recommendation applied.
4. Ask the assistants the agreed questions: per question, per assistant, the brand named, answered without it, or no answer, with the denominator.
5. Propose the rivals with evidence; you keep, drop and add; the kept ones are measured one at a time.
6. Agree the questions worth winning: relevant rows only, gap first, what each answer rests on.

Everything the sequence turns up, pages to write, briefs, sources to be on, is filed in the backlog for your decision, or handed out at once if you said the team may start.

After the sequence: the queue

Work is ranked from what was measured, in three lanes of proposals, against your strategy mix, with a recommendation and a confidence. The next piece is chosen without handing the same articles over twice. On every weekday review the lead says where the effort goes and why.

What a site needs

At the site level: the root files an assistant’s crawler reads, the identity pages, titles, headings and links, speed. At the page level, for editorial pages: what an assistant needs to lift an answer from it. A finding names its pages, up to twelve, and whose morning it belongs in: server and configuration to a developer; titles, headings, links and the identity pages to whoever edits the site. Never a grade, never a percentage of “health”: a finding is a consequence and a fix.

What gets a page into an answer

A page is quoted when it answers a question the category asks, in a shape an assistant can lift, with an original number or a clear claim, from a site the assistant trusts. The niche is the list of those questions: who is on the answer today, and what the answer rests on, as a map of where the category is decided.

The writing standard

Every piece, a page, a research piece, a brief for a developer, an llms.txt, is written to the same standard: the shape every piece takes; quotable by an assistant; not a machine; pictures where they carry meaning; where it goes on the site; and for a developer’s brief, what to do exactly. A draft is scored before it is handed over and fixed on its heaviest missing lever; the Fact-checker reads it last, tracing every figure and every source.

A repeat check

On a scheduled re-check, the analyst leads with what changed since the last one, losses first, and re-measures only what is past its cadence or what you asked to check fresh.