Skip to content
Working with an agency

What GEO agency onboarding actually looks like

The first 30 days with a GEO agency, phase by phase: the baseline audit, the access you'll need to grant, and what your first report should contain.

By Jeremy ZavaAI-assisted, reviewed by Ibrahem Helal6 min read

A solid GEO agency onboarding runs three phases inside the first 30 days: a baseline audit of your current citations and crawler access, fixes to your highest-intent pages, and the start of third-party outreach. You should get a written baseline within two weeks and a first tracked report by day 30.

This is the practical follow-through to how to choose a GEO agency: once you've picked someone, here is what the first month with them should actually look like.

On this page

What should happen before you even sign the contract?

The proposal and the RFP stage, if you ran one, should already have named the engines in scope, the third-party targets and the pricing — see what to put in a GEO RFP for that groundwork. What should not happen before signing is any real technical work; a vendor doing free fixes before a contract exists is either underpricing the relationship or trying to create a sense of obligation before you've committed to anything.

What is reasonable before signing is a short discovery call to confirm scope: which domains and subdomains are in play, whether you have an existing SEO vendor doing adjacent work, and who on your side will own access requests once the contract starts. That last point matters more than it sounds — naming the access owner before day one, rather than during it, is the single biggest predictor of whether week one actually finishes on schedule.

You should also leave that call with a written scope document, not just a verbal understanding. It does not need to be long: the engines in scope, the pages that will be prioritized first, and the cadence of reporting are enough to hold both sides to the same expectations once the actual work starts.

What happens in week one?

Three things, in roughly this order. First, a crawler access check: confirming GPTBot, ClaudeBot, PerplexityBot and Google's AI-related crawlers can actually reach your site, since a firewall rule can block them with no visible error [1][2]. Second, the prompt set: twenty to forty real buyer questions, sourced from your sales team or support tickets rather than a keyword tool, which becomes the fixed measurement for the rest of the engagement. Third, access — more on exactly what below.

Who should actually be on the kickoff call?

Fewer people than most companies bring, and the right ones. You need whoever owns the CMS or can grant developer access, since page fixes are worthless if they sit in an approval queue. You need whoever can speak to engineering about firewall and CDN rules, since that is where crawler blocks usually hide. And you need one person with the authority to approve the prompt set and the third-party pitch list, so the agency isn't waiting on a committee to sign off on twenty questions.

What you don't need in that first call is a large stakeholder group reviewing strategy from scratch. Strategy questions belong in the proposal stage, before signing; the kickoff call is about execution logistics — who grants what, and how fast.

How is your AI visibility baseline actually measured?

  1. Confirm crawler access

    Test with each crawler's real user agent and confirm a 200 response and full HTML, not just a robots.txt read.

  2. Run the fixed prompt set once

    Log which URLs each engine cites today, before anything changes — this single run is the number everything later gets compared to.

  3. Audit your highest-intent pages

    Pricing, comparison and specification pages first, since these are the pages most likely to be quoted if they answer clearly.

  4. List the third-party pages already citing competitors

    These become the first outreach targets — you're not starting from zero, you're starting from a known list.

What access does the agency actually need from you?

Search Console and Bing Webmaster Tools access, so citation-adjacent data can be cross-checked against a source you also control, not only the agency's own tracker [3]. CMS or developer access, or a clear process for submitting page changes, since fixes that sit in a queue for weeks are fixes that aren't happening. And visibility into your CDN or firewall rules, or a fast path to your engineering team, because that is where crawler blocks usually live and it's rarely somewhere a marketing-only login can reach.

Day 1 versus day 30, on a normal onboarding
Day 1Day 30
Crawler status unknownCrawler status confirmed and any blocks fixed
No fixed prompt set20–40 buyer questions tracked weekly
No documented baselineA written baseline report across all engines in scope
Zero third-party pitches sentFirst round of outreach sent to known gatekeeper pages
Highest-intent pages unreviewedA prioritized list of page fixes, some already shipped

What actually slows an onboarding down?

Access delays, almost every time. A CMS login that takes two weeks to provision, or a firewall change that needs a ticket nobody prioritizes, can stall real work while the calendar keeps moving. The fix is naming an internal owner on day one — someone who can approve access requests same-day — rather than routing everything through a shared inbox. The second most common delay is an unclear prompt set: if nobody on your side can commit to twenty real buyer questions in week one, the baseline slips and so does everything measured against it.

A third, quieter cause is a second vendor working the same pages without anyone coordinating. If your existing SEO agency is also touching the pages the GEO agency wants to rewrite, both can end up waiting on each other, or worse, publishing conflicting changes in the same week. Naming that overlap explicitly in the kickoff call, and agreeing on who edits what, avoids a stall that neither vendor will notice until a page has been changed twice in opposite directions.

What does a 60- and 90-day check-in look like?

By day 60, expect the first evidence of third-party movement — a reply from a site you pitched, even if the placement itself hasn't gone live yet, since these relationships take longer than a month to close. By day 90, expect at least one real placement or a clear account of why not, plus a citation-rate trend across three monthly readings rather than one. Three readings is the minimum that turns a single data point into an actual trend, given how much a single week's numbers can move on their own [4].

If 90 days pass with a flat baseline, no third-party replies and no explanation, that is a fair moment to ask hard questions about the engagement — not because 90 days guarantees results, but because it is long enough that "too early to tell" stops being a satisfying answer.

What should your first report actually contain?

The baseline numbers from week one, restated next to whatever moved by day 30 — crawler fixes shipped, pages rewritten, outreach sent and any early replies. It should name specific URLs, not a blended visibility score, and it should say plainly where nothing has changed yet, since third-party placements typically take longer than 30 days to land. How long GEO takes covers realistic timing in more depth, and what a GEO retainer includes covers what month two and three should add on top of this first report.

Read the report for what it does not say as much as what it does. A report that only lists activity — "sent twelve pitches, rewrote four pages" — without connecting any of it back to the baseline numbers is telling you the agency was busy, not that anything moved. A report worth reading ties every line of activity back to a number from week one, even when that number is unchanged, because an honest "no movement yet, and here's why" is more useful at day 30 than a summary that skips the comparison entirely.

The format matters less than the habit. Some agencies send a slide deck, others a shared spreadsheet updated weekly instead of a fixed monthly document — either is fine as long as you can trace every claim back to a specific URL and a specific date. What matters is that the report exists at all, on a predictable schedule, and that it reads the same way in month six as it did in month one.

Questions people also ask

  • How long should GEO onboarding take?

    Roughly 30 days for the baseline audit, initial fixes and the first tracked report. Anything under two weeks usually means the baseline was rushed or skipped; anything past six weeks with no report is worth asking about directly.

  • What access does a GEO agency actually need?

    Search Console and Bing Webmaster Tools access, CMS or developer access to fix pages and technical issues, and visibility into your CDN or firewall settings to check for crawler blocks. Analytics access is useful but secondary.

  • What's the very first thing that should happen?

    A crawler access check — confirming the major AI crawlers can actually reach your site. It costs nothing, takes under a day, and a silent block here undermines every other part of the engagement until it's fixed.

  • Should onboarding include a written baseline report?

    Yes. Without a documented starting point — current citation rate, crawler status, page issues found — you have no way to judge whether month three actually improved on month one.

Sources

  1. AI features and your website — Google Search Central (checked 22 September 2026)
  2. Overview of OpenAI crawlers — OpenAI (checked 22 September 2026)
  3. Introducing AI Performance in Bing Webmaster Tools — Bing Webmaster Blog (checked 22 September 2026)
  4. I tracked 13,184 AI citations across 4 LLMs: here's what I learned — Orbit Media Studios (checked 22 September 2026)

Want to see what your first 30 days would actually include?

Book a free 30-minute consultation with Ibrahem and Jeremy. We look at your site while we talk, and you leave with a prioritised plan.

Who wrote this

Jeremy ZavaCo-founder

Leads growth and client partnerships.

Works on: CRO, SEO

Reviewed before publishing by Ibrahem Helal, Co-founder.

Posts by Jeremy Zava

More on this