The first 90 days of a GEO programme should produce a clear baseline, useful page improvements and a repeatable way to review the evidence. Ninety days is a planning window, not a guarantee that AI platforms will cite your business or send a particular number of leads.
The most useful pilot starts small: one market, one service or product category and a manageable group of buyer questions. That makes it easier to understand what was changed and whether the work deserves further investment.
Days 1–15: define the question and baseline
Interview the people who handle sales and product questions. Identify what prospects need to know before choosing you: suitability, price, alternatives, integration, service coverage or implementation effort.
Map those questions to existing pages. Record what is accurate, missing, outdated or difficult to find. Agree which claims require product, legal or commercial approval before writing begins.
Create a repeatable observation log. It should retain the exact prompt, platform, date, relevant settings, answer, cited URLs and brand context. A mention, a recommendation and a linked citation should have distinct labels. Record failed collections separately.
Deliverables at this stage are a question map, page inventory, observation baseline and agreed measurement definitions. An attractive dashboard without those foundations is difficult to interpret later.
Days 16–30: prioritise implementation
Rank the gaps by buyer relevance, likely usefulness and delivery effort. Fix inaccurate or incomplete information on important commercial pages before producing a large collection of generic AI articles.
Write a task list with the affected URL, proposed change, evidence source, owner and acceptance check. Include technical issues where they prevent important content from being discovered or understood. Use the GEO audit checklist to structure that review.
For a fictional B2B service firm, the first batch might clarify service coverage, explain quote factors and connect an implementation guide to the relevant enquiry page. Those are bounded changes the team can publish and verify.
Days 31–60: publish and inspect the result
Move approved changes into production. Check the public page, internal links, forms and relevant metadata. Keep a publication log so the next review can distinguish planned work from completed work.
Bring supporting material into the customer journey: product documentation, carefully sourced comparisons, useful examples and genuine evidence. If a proposed article repeats an existing page’s intent, improve that page instead of creating another near-duplicate URL.
Continue the same observation routine. Do not replace difficult prompts with easier ones to make the score look better. If the sample needs to evolve, preserve a stable comparison group and label the new questions separately.
Days 61–90: decide what to continue
Review implementation, visibility and commercial outcomes as separate layers.
| Layer | Evidence | Decision |
|---|---|---|
| Delivery | Published URLs and verification log | Was the agreed assignment completed? |
| AI observations | Same-question answers, mentions and citations | Is relevant visibility changing within the sample? |
| Website acquisition | Identifiable referrals and relevant search visits | Are useful visitors reaching the improved pages? |
| Commercial outcomes | Qualified enquiries and CRM stages | Is the demand suitable for the business? |
Where observations remain sparse, document the uncertainty. Look for repeated patterns rather than treating one favourable answer as a durable result. Review other changes that could affect traffic, including launches, campaigns and seasonality.
What should trigger a change of plan?
If implementation is blocked, solve ownership or capacity before increasing content volume. If the questions are irrelevant to buyers, revise the research method and establish a new baseline. If visitors arrive but do not understand the offer, review the landing page and qualification journey.
If the work is complete but there is too little data, extend observation with a defined cost and review point. Continuing indefinitely without a decision framework is as unhelpful as abandoning a useful programme because a tiny sample fluctuated.
What the day-90 report should contain
Keep it concise: original objective, work completed, evidence collected, limitations and the next scoped decision. Include the underlying logs for anyone who needs to check the conclusion.
For expectations about timing, read how long GEO takes to show results. For funding the assignment, use the GEO budget guide. This article is an operating plan; the timing guide addresses how to interpret outcomes over time.
Ready to define a manageable pilot? Discuss your first market and buyer questions with Crescendo.
