What should happen in the first 90 days of an AEO engagement?

Last updated October 14, 2026. Written by the Third Coast AI team.

In the first 90 days of an AEO engagement you should see a baseline established over repeated runs by the end of week three, crawler and technical fixes shipped by week four, the fact-check corrections requested by week five and showing in retrieval-on answers by weeks eight to ten, the first decision-factor pages published to a gate by week eight, the entity record reconciled by week ten, and a twelfth-week report that shows leading indicators green, early corrections landed, and the visibility rate re-measured against the baseline with the run counts shown. What you should not expect by day 90 is a large move in the share of competitive prompts naming you, because that builds over two quarters as pages are indexed, mentions are earned and the runs compound.

This page is the schedule we work to and the one we would hold any vendor to, week by week, with what you have to supply at each point. It assumes a single brand in a single market; each additional market or brand adds about a week to the first phase. The page on how long AEO takes explains the timing beyond day 90.

What happens in weeks one to three?

The first three weeks are the audit, and they end in a baseline and a plan. Week one is intake and the prompt set: about two hours of your time on a call where we learn how buyers ask, who the personas are and which topics you need to own, and the rest of the week writing 60 to 120 prompts across the visibility, decision-factor, competitor and fact-check groups and getting your sign-off on them. Week two is the runs with retrieval on in ChatGPT and Gemini plus AI Overviews, AI Mode, Perplexity and Copilot, each prompt several times, because SparkToro and Gumshoe.ai found less than a 1 in 100 chance of the same brand list repeating on a single run; from those runs we capture the hidden searches the engines run behind each question and the decision factors in the engines' own words, log every wrong fact and its cited source, and read the pages cited for each topic. Week three is the proof audit of your pages against the factors, the citation reads written up as briefs, and the readout with a ranked 90-day plan. The what-to-expect guide lists the deliverables in detail.

Which fixes should land first, and why?

Crawler access, because nothing downstream works without it. Cyrus Shepard's May 2026 synthesis of 54 studies scores crawler accessibility 9.5 out of 10, the highest of 23 factors, and OpenAI's crawler documentation says a site that disallows OAI-SearchBot will not be shown in ChatGPT search answers and that a robots.txt change takes about 24 hours to register. The fix is often a CDN setting rather than robots.txt, and it is the one week-four item that can change the baseline on its own. Second are the fact corrections, because they are specific, traceable and fast; the correction guide describes the method, and the first corrected answers usually appear around weeks eight to ten. Third are the first decision-factor pages, chosen from the proof audit as the topics where you are absent and the gap is widest, written to briefs modeled on the cited pages and passed through the gate before publishing. Entity work runs alongside from week seven because it is slow to take effect and cheap to start.

How many pages should ship in the first quarter?

Fewer than most proposals promise. Four to eight pages through a real gate is a strong first quarter for one brand, because each page needs a brief from the citation reads, a subject-matter reviewer on your side, and the structural elements the research rewards: Aggarwal and colleagues at Princeton measured quotations, statistics and cited sources lifting a page's visibility in generated answers by about 41%, 33% and 28%, and Zhang and Yao found top-quartile heading count cited at 12.5 times the bottom quartile, which is work per page rather than volume. A vendor promising thirty pages in a quarter is either generating them or skipping the reviewer, and Glenn Gabe documented a site that lost visibility after publishing 850,000 AI-generated pages. The GEO content program describes how we pace it.

What should the week-twelve report show?

• Every AI crawler fetching the main pages, with the agent-analytics or log evidence, and Bing Webmaster Tools AI Performance verified and reporting.
• The fact-check log with each wrong claim, its source, the fix, and the first run in which the corrected answer appeared, by engine.
• The pages shipped, their gate results, and whether each has been crawled and cited yet.
• The entity record: profiles reconciled, Wikidata item live, schema sameAs closed, knowledge panel claimed where one exists.
• The re-run of the full prompt set against the baseline, with run counts, reported by engine, and an honest note on which changes are inside the variance.
• Outreach started: which cited sources were approached and any mentions earned.
• The plan for the second quarter, with what you need to supply.

The visibility rate at week twelve will usually show small movement concentrated in the prompts whose pages were rewritten and in the fact-check group, and it may show nothing outside the variance yet, which is the expected result and not a failure. Ahrefs measured only 38% of AI Overview citations coming from Google's top ten in March 2026, so being indexed is not being cited, and the citations follow the mentions and the re-runs over the following months. The KPI guide explains why the leading indicators are the ones to read at this stage.

Can the engagement stop at day 90?

Yes, and it should be built so that it can. By week twelve you should hold the prompt set, the raw answers, the briefs, the gate, the published pages and the entity records, and the running can move inside with a tool subscription in your name and a quarterly outside re-read; the in-house versus agency guide covers that handover. If you continue, the second quarter is where the visibility rate starts to move, because the pages have been indexed, the corrections have settled and the earned mentions begin to appear in citations, and six months from the baseline is the honest horizon for calling a trend. Either way, the first 90 days should leave you with a system you understand and a report you could explain without us in the room.

Frequently asked questions