Short answer
A 90-day GEO plan has four phases. Weeks 1 and 2: write your fact sheet, build a blind question set and take a baseline. Weeks 3 and 4: fix technical access and your core pages. Weeks 5 to 10: publish answer-first pages for your top gaps and do honest outreach to the lists, directories and review sites AI apps cite. Weeks 11 to 13: re-measure, compare treated questions against a control group, and plan the next quarter from what the numbers show.
On this page
Before you start
This plan assumes a small team: one person owning GEO for a few hours a week, plus help from whoever writes content and manages the site. Stretch it if you have less time. Do not skip the measurement weeks.
| Phase | Weeks | Goal |
|---|---|---|
| Baseline | 1 to 2 | Know where you stand, honestly |
| Foundations | 3 to 4 | Remove blockers and fix core pages |
| Build | 5 to 10 | Close the biggest gaps on and off your site |
| Test and plan | 11 to 13 | Find out what worked and decide what is next |
Weeks 1 and 2: baseline
- Week 1. Write your entity sentence: who you are, what you sell, who it is for.
- Week 1. Build a fact sheet: products, prices with GST treatment and date, locations, founding year, key features. Confirm every line.
- Week 1. Collect 30 to 60 buyer questions from sales and support. Remove your brand name from all of them.
- Week 2. Tag each question by cluster and tier. Pick your Tier 1 set.
- Week 2. Choose a control group: similar questions you will not work on this quarter.
- Week 2. Run a baseline: each question, on each AI app, several times, with search on. Record mentions, citations, position, competitors and sources.
Weeks 3 and 4: foundations
- Week 3. Check robots.txt, CDN and firewall for blocked search agents. Fix them.
- Week 3. Confirm key pages render their text in plain HTML.
- Week 3. Update your sitemap and submit it.
- Week 3. Add Organization JSON-LD with your entity sentence.
- Week 4. Rewrite your About page around the entity sentence and fact sheet.
- Week 4. Rewrite your pricing page: rupees, GST treatment, billing period, date.
- Week 4. Claim and correct your main directory profiles so they match the fact sheet.
Weeks 5 to 10: build
Use the baseline to pick your targets. For each Tier 1 question where you are missing, decide the cause:
| Cause | Fix | Where it goes in the plan |
|---|---|---|
| No page answers the question | Publish a citation-first page | Content track |
| A page exists but buries the answer | Rewrite with an answer box and clear facts | Content track |
| AI apps cite lists you are not on | Honest outreach to authors | Outreach track |
| AI apps cite directories or reviews | Complete profiles, ask real customers for reviews | Outreach track |
| AI apps describe you wrongly | Fix facts on your site and ask others to correct theirs | Both tracks |
A realistic weekly rhythm for a small team:
- Content track: one or two new or rewritten pages a week, each passing the ten-point checklist.
- Outreach track: a handful of well-researched pitches or profile updates a week.
- Review requests: a steady flow to real customers, with no incentives.
- Internal links: link each new page from two or three related pages.
Weeks 11 to 13: test and plan
- Week 11. Re-run every question on every AI app with the same settings as the baseline.
- Week 11. For each AI app, compute the mention rate with a 95% range for treated and control questions.
- Week 12. Use a two-proportion test to compare baseline and now. Call a change real only if it clears the 95% level.
- Week 12. Compare treated with control. If both moved together, look for an outside cause.
- Week 12. Read the new source lists. Did the pages you pitched start naming you? Did your own pages start being cited?
- Week 13. Write a one-page review: what moved, what did not, what you learned.
- Week 13. Plan the next quarter: keep what worked, drop what did not, pick new Tier 1 gaps.
Setting expectations
Some changes show up only after pages are crawled and indexed, and answers that come from model memory change only when models are updated. A quarter may show clear movement on some AI apps and none on others. That is normal. Report honestly, with ranges, and keep going.
- Do report: rates with ranges, treated vs control, source changes.
- Do not report: one screenshot, a single lucky answer, or a change that did not beat the noise.
How we help
We follow this same loop in AI Knows Us: baseline, gap diagnosis, drafted pages and outreach, and tiered re-measurement with control groups, so a small team can run the plan in a few hours a week.
Frequently asked questions
Can I do this in less than 90 days?
You can do the work faster, but you still need time for pages to be crawled and for enough checks to build a reliable rate. Short tests often end in noise.
How many questions should I track?
Enough to cover your main buyer jobs. Keep the Tier 1 set small enough to check often, and the rest on a slower schedule.
What if nothing changes by week 13?
Check the basics first: were pages crawled, did AI apps search, were the control and treated groups similar? Then look at whether the gaps were on third-party pages you have not reached yet.
Should I work on every AI app at once?
Measure all of them, but prioritise fixes that help several AI apps, such as crawlability, clear pages and third-party presence.
Who should own GEO in a small company?
Usually the person who owns content or marketing, with help from whoever manages the website.
Sources
Last verified: 17 September 2026. Facts on this page come from the public documents below. Brands such as Tallybook are fictional and used only as examples.