A startup AEO plan is a 90-day sequence of fixes and content that makes your company easy for AI answer engines to find, understand and cite. For a team with one marketer and no budget, the order that works is: make sure crawlers can reach you, record a baseline of real buyer prompts, rewrite the few pages that matter, publish the pricing and comparison content buyers ask for, then measure. As of 14 October 2026, nobody can promise citations on a schedule, so this plan is built around controllable inputs and honest checkpoints.
Key Takeaways
- Start with access and baseline, not content. A blocked crawler or a missing page makes everything after it wasted.
- Google says there are no special requirements to appear in AI Overviews or AI Mode. Ordinary SEO basics (indexable, snippet-eligible, helpful pages) are the foundation.
- OpenAI documents separate crawlers for search (OAI-SearchBot) and training (GPTBot). Blocking the wrong one removes you from ChatGPT search.
- Timelines in AEO guides ("citations in 2 to 4 weeks") are mostly unsourced. Plan for weeks to see movement on narrow prompts and quarters for category-level presence.
- Report against a baseline of prompts and citations, not only sessions. AI-referred traffic is often small.
- The plan assumes one part-time marketer, about 6 to 8 hours a week. If you have less, drop the competitor page and the third-party work first.
Before you start: set honest expectations
Three facts shape this plan.
- Google's AI features run on ordinary Search eligibility. Google's AI features documentation says a page must be indexed and snippet-eligible, and that there are "no additional requirements" or special schema for AI Overviews and AI Mode. Indexing and serving are not guaranteed.
- A new site needs time regardless. Google's SEO starter guide says some changes take hours and others several months, and suggests waiting a few weeks before judging.
- Most "90-day" claims are anecdotes. One vendor guide for startups says well-structured content may appear in AI answers within 2 to 4 weeks and consistent multi-platform presence takes 3 to 6 months, but gives no data. Another, a week-by-week plan for a cybersecurity vendor, is more cautious: it says entity authority takes quarters, and that traffic may not move much. I take the cautious view.
The plan at a glance
| Weeks | Phase | Output | Hours/week (approx.) |
|---|---|---|---|
| 1 to 2 | Access and baseline | Crawler audit, 30-prompt baseline, competitor list | 8 |
| 3 to 4 | Foundations | Entity consistency, Organization markup, page inventory | 6 |
| 5 to 8 | Key pages | Rewrite 5 pages; publish pricing and use-case pages | 8 |
| 9 to 10 | Buyer-question content | Comparison/alternatives page, FAQ from sales calls | 8 |
| 11 to 12 | Measure and decide | Re-run prompts, report vs baseline, next-quarter plan | 6 |
Weeks 1 to 2: access and baseline
Check that crawlers can reach you
Pull your robots.txt and your CDN or firewall bot rules. OpenAI's crawler documentation lists separate bots:
- OAI-SearchBot surfaces sites in ChatGPT's search features.
- GPTBot crawls content that may be used to train OpenAI's foundation models.
- ChatGPT-User visits pages when a user asks ChatGPT something.
Each setting is independent, so you can allow search while disallowing training. OpenAI says sites that opt out of OAI-SearchBot will not appear in ChatGPT search answers, and that search-related robots.txt changes can take about 24 hours to take effect. Decide your policy on purpose, write it down, and check the firewall as well as the file; a legacy bot rule is a common culprit, as the cybersecurity plan above also notes.
Confirm Google can index you: run site:yourdomain.com, inspect key URLs in Search Console, and make sure core content is in the HTML rather than loaded only by JavaScript.
Build a 30-prompt baseline
Write 30 prompts a real buyer would type: category ("best [tool type] for [use case]"), problem ("how do I [task]"), comparison ("[you] vs [competitor]") and brand ("is [you] legit"). Pull the wording from sales calls, support tickets and community threads, not from imagination.
Run each in the main engines you care about, on the same day, logged out where you can. Save the full answer, the cited URLs and whether you or a competitor appear. My method is in the 50-prompt tracking panel. Expect to appear in very few at the start; that is the point of a baseline.
Name your real competitors
List who AI answers actually mention, which may differ from who you consider competitors. These are your comparison and alternatives targets later.
Weeks 3 to 4: foundations
Make your entity consistent
Write one factual description of the company (what it is, who it is for, where it is based, founding year) and use it everywhere: homepage, About page, LinkedIn, Crunchbase, app listings, review sites. Inconsistent descriptions are a common reason answers describe the wrong category. My note on the entity home page goes deeper.
Add Organization markup where it fits
Google's Organization structured data guidance says to place it on your home page or one page that describes the organisation, and to include recommended properties such as name, url, logo and sameAs. It is not required for AI features; it is a way to state facts clearly.
Inventory pages
List every page that answers a buyer question: product, pricing, use cases, integrations, docs, FAQ. Mark each as missing, thin or fine. This becomes the weeks 5 to 10 work list.
Weeks 5 to 8: rewrite the pages that matter
Pick the five pages most likely to be read by a buyer: home, product, pricing, top use case and your best-performing blog post. For each:
- Open with a plain statement of what the product or page covers, naming the audience.
- Replace marketing adjectives with specifics: what it does, for whom, at what price, with what limits.
- Convert comparisons and plan details into HTML tables, not images.
- Attribute every statistic with source and date.
- Show a last-updated date.
Then publish what is missing. Priorities for a startup:
- A public pricing page with real numbers, or at least ranges and what drives them. A Sproutworth guide argues vague "contact us" pricing reduces eligibility for citations; the example it gives is a single anecdote, so treat it as a hypothesis. Buyers still want the number.
- One page per top use case, written as the answer to "how do I do X with [category]".
- Integration pages if you have integrations; see my notes on SaaS pricing and integration pages.
Weeks 9 to 10: buyer-question content
An honest alternatives or comparison page
Buyers ask AI engines to compare. A page that fairly compares you with two or three alternatives, with a table, criteria and a "who should choose which" section, gives engines something to quote. Be truthful about where competitors are stronger. See how SaaS alternatives pages work.
Turn sales calls into FAQs
Take the ten questions your sales or support team hears most. Answer each in 2 to 4 sentences on the relevant page, visible to readers. Add FAQ markup only if it matches visible text and you can maintain it.
Original evidence
If you have anonymised product data, a customer result you have permission to share or a test you ran, publish it with the method. Information that other pages do not have is the best reason to be cited; see information gain.
Weeks 11 to 12: measure and decide
- Re-run the same 30 prompts on the same engines. Compare against the baseline: appearances, citations, which page was cited, what was said about you.
- Check Search Console for impressions and clicks on the pages you changed. Google says clicks from AI features count in the Performance report under the Web search type, so you cannot separate them there.
- Check analytics for referrals from AI tools. Expect small numbers.
- Check server logs, if you can, for crawler visits to the new pages.
- Write a one-page report: what changed, what moved, what did not, and what you will do next quarter.
What to cut if you have less time
| If you have | Do | Skip |
|---|---|---|
| 2 hours a week | Crawler check, 15-prompt baseline, rewrite 2 pages | Alternatives page, third-party work |
| 5 hours a week | Above plus pricing page and 3 FAQs from sales calls | Original data project |
| 8+ hours a week | Full plan | Nothing, but keep weeks 11 to 12 |
Skip llms.txt and monitoring-tool purchases early on. In June 2025 Google's John Mueller wrote that "no AI system currently uses llms.txt", as reported by Search Engine Roundtable; that is a dated statement, so re-check before you decide. A spreadsheet is enough for 30 prompts.
Things that commonly go wrong
- Pricing and comparison pages stall in approval. Get leadership agreement in week 1, not week 8.
- Descriptions drift. A sales deck, LinkedIn and the website say three different things.
- Tracking noise. AI answers vary run to run. Use the same prompts, record dates, and look for patterns across several runs rather than single results.
- Overreading early wins. One citation after two weeks is an anecdote.
- Ignoring Google basics. If pages are not indexed, nothing else matters.
For my weekly routine once the plan is running, see my AEO weekly review.
FAQ
What is an AEO plan for a startup?
It is a time-boxed set of actions to make a young company easy for AI answer engines to find and cite: crawler access, consistent entity information, clear key pages, content that answers buyer questions and a measured baseline. It sits on top of ordinary SEO rather than replacing it.
Can a startup get cited by AI in 90 days?
Possibly for narrow prompts, but nobody can promise it. Some guides claim citations in 2 to 4 weeks, without data. Plan to see changes on specific prompts within the quarter and treat category-level presence as a longer project.
Do I need an llms.txt file?
In June 2025 Google's John Mueller said no AI system currently uses it (reported by Search Engine Roundtable), and the cybersecurity vendor plan I reviewed advises skipping it early. Re-check that before relying on it. If you add one, it should be a low-priority, low-effort task after the fundamentals.
Should I block GPTBot?
That is a policy choice. OpenAI says blocking GPTBot signals that content should not be used for training, while OAI-SearchBot controls ChatGPT search. If you want to appear in ChatGPT search, allow OAI-SearchBot even if you disallow GPTBot.
Do I need schema markup for AEO?
Google says no special structured data is needed for AI features. Organization markup is still a clean way to state facts about your company, and it must match the visible page.
What should I measure?
Measure your prompt set (appearances and citations), Search Console performance for changed pages, AI referral sessions and crawler visits in logs. Report against the baseline you recorded in weeks 1 and 2.
Is it worth doing without a budget?
Yes, the first 90 days need no tools: a spreadsheet, access to your site and a few hours a week. Paid monitoring can wait until you know which prompts matter.
What if my pricing is custom?
Publish what you can: starting prices, typical ranges, what drives the cost and what is included. Buyers and answer engines both benefit from specifics, and you can still invite a call for the final quote.
How is this different from an SEO plan?
It shares the foundations (indexable, helpful pages) but adds prompt-based tracking, extraction-friendly page structure, third-party and entity consistency, and reporting on citations as well as rankings.
When should I hire help?
When you have a baseline, a clear list of gaps and not enough hours to close them. An outside strategist is most useful for audits and prioritisation; execution can stay in-house.
Work with me
I'm Younus Fardeen, an organic growth, SEO and AEO strategist in India with 4+ years of marketing experience across edtech and startup clients. If you'd like me to look at your crawler setup, prompt baseline and key pages, see my work and reach me through the contact form at younusfardeen.in.