An SEO Agent Onboarding Checklist for a New Client Domain

The first week with a new client domain determines whether the engagement runs on evidence or on vibes. Skip the baseline and every future report opens with an apology; skip the boundaries and the first incident is a relationship event instead of a log line. The checklist below is written for engagements using agentic SEO tooling, and gates two through four apply unchanged to fully manual ones — the agent just makes the discipline non-optional.

The four gates

1. Access — GSC property verified (or delegated), analytics read access granted, the deploy path named and tested (CMS API? git? a person?), and credentials scoped per client in your secrets manager. Shared credentials across clients fail the evidence-separation test the first time anyone audits
2. Baseline — snapshot 30 days of GSC queries and pages, run the buyer-intent prompt set for the first time and store the results with dates, archive robots.txt, the sitemap, and key pages' current schema. This is the 'before' photo; it cannot be taken retroactively
3. Boundaries — the approval matrix signed (who approves which risk class), the do-not-touch list written and acknowledged (legal pages, active campaigns, the client team's own work-in-progress), and the escalation contact named for anything ambiguous
4. Rollback — revert a trivial test change through the real deploy path before shipping anything that matters. A rollback that's never been rehearsed is a rollback that fails during the incident that needed it

The checklist artifact

The failure example

The missing baseline: an agency ships six weeks of genuinely good work, the client asks 'what's changed since you started?', and the honest answer is a shrug with charts that begin week three. Rankings that improved from an unrecorded starting point improved from nowhere; the AI-visibility wins can't be distinguished from prompts that were already citing the client. The baseline gate costs one afternoon; its absence costs the engagement's entire evidentiary foundation, permanently — there is no retroactive 'before'.

Where Fenn fits, honestly

Fenn covers less of this than a tool that deploys would. Plans are set up with you after a demo, a Fenn measurement gives you the prompt-set first run for the baseline gate (export it to CSV and store it dated), and its review queue and activity log hold the approvals. It doesn't touch access or rollback, because Fenn doesn't change the site — you or your developer publish each approved fix. What it cannot do is the contractual half: the approval matrix and do-not-touch list are agreements between you and the client, and the rehearsed-rollback gate still deserves a human's eyes on the client's real deploy path. Tooling makes the checklist cheap; it doesn't make it optional.

FAQ

How long should onboarding take? — One focused week: access is days of client-side waiting (start it first), the baseline is an afternoon once access lands, boundaries are one meeting, rollback is an hour. Engagements that skip to 'quick wins' in week one pay the week back with interest at the first report or the first incident.

Does this apply to non-agentic engagements? — Gates two through four, identically — baselines, boundaries and rollback are evidence discipline, not agent safety. Gate one's credential scoping matters most when tooling acts on your behalf, but shared-credential audits embarrass manual agencies just as reliably.