A GEO strategy is a plan to get your brand mentioned, recommended and cited when buyers ask AI assistants about your category. In 90 days, a realistic version looks like this: build a prompt set and a baseline in the first three weeks, fix access and facts on your own pages next, then spend the middle of the quarter on third-party sources and content, and close with a re-measurement that tells you what moved.
Most teams don’t need a new department for this. They need an order of operations, an owner per workstream and deliverables that can be checked. This plan is written for a small B2B marketing team, with a CRM vendor as the example. It doesn’t promise rankings, since AI answers vary and no vendor publishes how it picks brands. It gives you a structured way to improve your odds and to know whether it worked.
Key takeaways
- A 90-day GEO strategy runs in five phases: prompts and baseline, technical and factual fixes, third-party sources, content, and measurement.
- The baseline audit comes first, because every later decision depends on knowing where you’re absent and which sources assistants cite.
- Each workstream needs one owner and one deliverable, or GEO turns into a vague side project.
- Third-party sources and your own content run in parallel in the middle weeks, since assistants often cite both.
- Most of the work is SEO, content and PR discipline pointed at new questions, which matches Google’s own guidance that optimizing for its AI features is still SEO.
- Judge results on trends across repeated runs at day 90, not on a single answer.
Before you start: scope, owners and expectations
Scope. Pick one product line, one market and one language. A focused quarter beats a scattered one.
Owners. Name a GEO lead (often the SEO or content lead) who owns the prompt set, baseline and reporting. Web or dev, product marketing, PR and content each own one workstream below.
Expectations. Google’s guide to optimizing for generative AI features says “optimizing for generative AI search is optimizing for the search experience, and thus still SEO,” and tells site owners to check third-party “AEO” or “GEO” advice against its official guidance. This plan contains no tricks, only work that should also help your search presence and your buyers.
The 90-day GEO plan at a glance
| Weeks | Phase | Owner | Deliverable |
|---|---|---|---|
| 1 | Prompt set | GEO lead | 15 to 30 frozen prompts, tagged by intent |
| 2 to 3 | Baseline audit | GEO lead | Baseline sheet, share of voice, cited domains, gap list |
| 3 to 4 | Technical and factual fixes | Web and product marketing | Crawler access confirmed, key pages rewritten with clear facts |
| 5 to 9 | Third-party sources | PR and partnerships | Outreach list worked, listings and corrections done |
| 5 to 10 | Content | Content | Comparison, use-case and fan-out pages shipped |
| 11 to 12 | Measurement | GEO lead | Re-run on the same setup, before/after report |
| 13 | Review | Whole team | Next-quarter plan |
Weeks 1 to 3: prompts and baseline audit
Week 1: build the prompt set
Write the questions buyers ask an assistant before they buy, in their words: category discovery (“best CRM for a 20-person startup”), use case (“CRM with email sequences for a small B2B agency”), comparisons (“HubSpot vs Pipedrive for a small sales team”), problem-led questions and a few branded checks. Pull wording from sales calls, support tickets and Search Console. Our guide to choosing prompts to track covers selection and tagging.
Deliverable: a frozen list of 15 to 30 prompts, each tagged by intent, segment and priority.
Weeks 2 to 3: run the baseline
Run every prompt more than once in each assistant your buyers use, and log mentions, position, competitors named and cited sources. Note which accounts and settings you used, so the day-90 re-run is comparable. The AI visibility audit template gives you the columns and scoring.
Deliverable: a baseline sheet with visibility rate and share of voice per assistant, the most cited domains, and a gap list ranked by revenue impact and effort.
Weeks 3 to 4: technical access and factual fixes
Start with what blocks you completely. Check robots.txt, CDN and bot-protection rules for the crawlers that matter. OpenAI’s crawler documentation states that sites opted out of OAI-SearchBot won’t be shown in ChatGPT search answers, and our AI crawlers and robots.txt guide lists the main user agents. For Google, the AI optimization guide says pages must be indexed and eligible to show with a snippet.
Then fix the facts on your key pages (homepage, pricing, product and comparison pages): who the product is for, what it does, what it costs and what it integrates with, as text. If assistants repeated wrong information in the baseline, trace it to the sources they cited and correct it there too.
Owners: web or dev for access, product marketing for facts. Deliverable: a short changelog of what was fixed and when, so you can connect later changes in answers to specific work.
Weeks 5 to 9: third-party sources
For “best X” and comparison prompts, assistants often cite review platforms, roundups, media and community threads. Your baseline already lists them. Sort the most cited domains into four groups: your pages, competitor pages, third-party pages that mention you, third-party pages that don’t.
The last group is the outreach list. Typical actions:
- Complete and update review site profiles, and ask real customers for reviews through your normal channels.
- Contact authors of cited roundups with accurate, current information. Some will update; many won’t.
- Correct errors in third-party pages that describe you wrongly.
- Join communities where your buyers ask questions, as yourself and with disclosure.
Avoid shortcuts: Google’s guide says seeking inauthentic mentions isn’t helpful and its spam systems block it. Our guide on getting cited through review sites goes deeper.
Owner: PR, partnerships or community. Deliverable: the outreach list with status per domain (listed, updated, corrected, declined).
Weeks 5 to 10: content
Use two inputs from the baseline: prompts where you’re absent, and the web searches assistants ran. ChatGPT, for example, may turn one prompt into several searches, and pages that don’t rank for them are unlikely to be read.
Prioritize:
- Comparison and alternatives pages that are fair and specific about when each option fits.
- Use-case pages for the segments in your prompt set.
- Answers to recurring sub-queries your site doesn’t cover yet.
Write them for buyers first: answer early, state facts in self-contained sentences, use tables, date the claims that change and show sources. Our guide on writing content AI assistants cite has before/after examples. Don’t spin up dozens of near-identical pages for query variations: Google’s guide says that can fall under its scaled content abuse policy.
Owner: content, with product marketing reviewing facts. Deliverable: a fixed number of pages shipped (for a small team, often four to eight in this window), each mapped to the prompts it targets.
Weeks 11 to 13: measurement and review
Re-run the same prompts with the same accounts and settings as the baseline. Compare visibility rate, share of voice per assistant, average position and cited domains, prompt by prompt. Then look at downstream signals: AI referral traffic in analytics, and Google’s Generative AI performance report in Search Console, which the AI optimization guide points to for measuring performance in its AI features. Our guide to measuring GEO ROI explains how to present these without overclaiming.
In week 13, decide what to keep, stop and add: a second segment, monthly tracking, another outreach cycle on newly cited domains. Our GEO checklist works well as the recurring task list.
Deliverable: a before/after report with the changelog attached, and the next quarter’s plan.
Run the measurement side with Citegram
The baseline in weeks 2 to 3 and the re-run in weeks 11 to 12 are the repetitive parts. Citegram, a Chrome extension, handles them:
- Add your prompt set and competitors. The free plan covers 1 prompt and 1 competitor; Pro covers unlimited prompts and competitors.
- Run them in the real assistants. Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, with the accounts already signed in to your browser. Runs count toward your own plan usage, and first results arrive in about 10 minutes.
- Get the baseline numbers. For each answer it records whether your brand was mentioned, its position and the cited sources.
- Feed the content and outreach phases. It records ChatGPT’s web searches and whether a Google AI Overview appeared and which sources it cited; the most cited domains report becomes your outreach list.
- Compare at day 90. Share of voice per brand and per assistant, with trends over repeated runs, shows what changed.
FAQ
What is a GEO strategy?
A GEO strategy is a plan to improve how often AI assistants mention, recommend and cite your brand for the questions buyers ask. It typically combines a tracked prompt set, technical access, clear facts on your own pages, presence on the third-party sources assistants cite, targeted content and regular measurement.
How long does GEO take to show results?
It varies. Some fixes, like unblocking a crawler or correcting a cited third-party page, can show up in answers within weeks, while content and reputation work takes longer. A 90-day plan is long enough to complete one full cycle and compare against a baseline, but no one can guarantee a specific result.
Who should own GEO in a marketing team?
Usually the SEO or content lead, because the work overlaps heavily with search. They own the prompt set, baseline and reporting, while web, product marketing, PR and content each own one workstream with a clear deliverable.
Start your baseline this week
Week 1 only needs a prompt list. To run it across assistants, the free plan tracks 1 prompt and 1 competitor, forever, and Pro adds unlimited prompts and competitors for $10 per seat per month. See pricing.
ChatGPT
