Phase 1, step 1 (days 1–7): run the audit and fix the technical fails
Connect the site's CMS or repository and the search console in Setup with scoped credentials, then run the nine-check audit: content and schema present without JavaScript, schema honest and validating, llms.txt current, answer-first openings, self-contained sections, consistent entities, verifiable claims, fan-out coverage, and a measurement baseline. Fix the technical fails first; they are fast and they unblock everything else.
Review the Brand Brain the wizard pre-filled: what the organization does, for whom, its products, its voice, and the claims it never makes. Every draft will be built on it.
Phase 1, step 2 (days 7–14): write the prompt set and sample the baseline
With the Search Visibility Digital FTE, write the prompt set: the primary questions buyers ask in your category and their fan-out sub-queries, five to nine per question. Sample it once across every engine that matters, recording citation presence, position, competing sources, and answer text with the date. Export the console data for the same period. This is the baseline, and it is the only thing that makes day 90 a comparison rather than a feeling.
Unify entity names across the site and the listings you control in the same fortnight, because the FTE's pages will use the canonical names and the rest of the site should agree.
Phase 2 (days 15–45): rewrite the pages that already rank
The FTE proposes, with a worth-it verdict each, the existing pages that rank for target sub-queries but fail the structure checks: buried answers, dependent sections, unsourced claims. Approve the verdicts, and the drafts arrive through the fact gate with every claim sourced. Review for judgement and tone, send back with reasons, and publish through the gated route. These pages are closest to citation because the ranking foundation already exists.
Keep sampling weekly. Some rewritten pages will show citation movement within the phase; most take longer, and the record is what shows which.
- Verdict approved: the page is worth rewriting, with reasons
- Draft through the fact gate: sources on every claim
- Approval: judgement and tone, never fact-checking
- Publish: CMS or pull request, always gated
Phase 3, step 1 (days 46–75): build the cluster from the gaps
From the sampled record, the FTE proposes topics for the sub-queries nobody on the site answers, each with a worth-it verdict; most verdicts are no, and the yeses are the content plan. Each new piece is written to the discipline, carries at least one thing the index lacks, links three related pieces, and names the same entities the same way. Voice-ready FAQs go on every piece; where a real video exists, chapters and a transcript.
Read the send-back reasons weekly; by this phase they should be few and about tone, because the Brand Brain has absorbed the earlier ones.
Phase 3, step 2 (days 76–90): the first quarterly comparison
Hold the quarterly review from the FTE's own report: citation presence per engine with sample sizes against the baseline, console impressions and clicks for the target queries, referral and branded trends as corroboration, pages published that passed the fact gate, approval rate with reasons, and usage. Everything not sampled or not computable is labelled unmeasured. Decide the three things: targets, remit, and whether any class has earned autonomy.
Publishing should not be a candidate. The FTE's value has been the verdicts, the sourced drafts, and the measurement; unattended publishing adds nothing the organization needs.
What does done look like at day 90?
Done is a site that passes the technical checks, a prompt-set record with three months of dated samples, the ranking pages rewritten to the discipline, a cluster started from evidenced gaps, an entity named consistently, and a quarterly report that says what moved, what did not, and what is unmeasured. The marketing lead spent minutes a day on verdicts and drafts and an hour a month on the review, and can show anyone the record.
It is not a promise of rankings. It is a managed loop with a record, which is the thing that compounds.
What should you refuse to do in the first quarter?
Refuse to promote publishing. Refuse to change the prompt set mid-quarter, because samples must be comparable. Refuse to add a source after the fact to pass a claim; send the draft back. Refuse to estimate what the performance view marks unmeasured. And refuse any tactic that promises a citation, because none is documented and several violate engine guidelines.
The refusals keep the record clean, and the record is what the second quarter is judged against.