# The 90-day AI search plan for a B2B marketing team of one

> A week-by-week 90-day plan for a solo B2B marketer: baseline, fix access, publish answer pages, earn list placements and report honestly.

Source: https://www.fame.so/post/90-day-ai-search-plan  ·  Fame Team  ·  November 3, 2026

## Key takeaways

- Days 1-14: set a baseline, fix crawl access and map 20 buyer questions.
- Days 15-60: improve or publish about one answer page a week, starting closest to revenue.
- Days 61-90: pursue third-party lists, refresh what is working and report trends.
- Budget about five hours a week, and expect noise. Ninety days is a test, not a guarantee.

A solo B2B marketer can run AI search in 90 days with about five hours a week: spend two weeks on a baseline and crawl fixes, six weeks publishing one answer page weekly, and the final month on third-party lists, refreshes and reporting. Results are not guaranteed, but you will know what moves and why.

## What makes this plan realistic for one person?

It rests on four choices, each tied to available evidence.

- Small fixed panel. Twenty buyer questions you actually check beat 200 you do not. [Ahrefs' stability study](https://ahrefs.com/blog/how-stable-are-ai-citations/) shows single checks are unreliable, so consistency matters most.
- Fundamentals first. [Google says](https://developers.google.com/search/docs/appearance/ai-features) its AI features have no additional requirements, and [Ahrefs reports](https://ahrefs.com/blog/how-to-rank-in-ai-overviews/) 76% of AI Overview citations also rank in the top 10.
- Depth over volume. Ahrefs found word count correlates near zero with citations, and [Google warns](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) against using automation mainly to manipulate rankings.
- Own site and third parties. [Vendor studies](https://www.tryanalyze.ai/blog/state-of-ai-search-in-b2b-2026) [disagree](https://wellows.com/blog/ai-visibility-for-b2b-marketing-agencies/) on the mix, so do both in sensible proportion.

The plan's timings and hour counts are planning assumptions rather than research findings. Adjust them to your site.

## What do you do in days 1 to 14?

The goal is to learn where you stand and remove blockers.

Week 1:

- Write 20 buyer questions from sales calls, support tickets, reviews and Search Console. Keep at least 15 unbranded. [Analyze AI found](https://www.tryanalyze.ai/blog/state-of-ai-search-in-b2b-2026) that naming the brand adds about 60 percentage points to mention rates.
- Group them: category discovery, problem, comparison and fact-check.
- Set up the [tracking spreadsheet](https://www.fame.so/post/track-ai-visibility).
- Run the first baseline in ChatGPT, Perplexity, Claude and Gemini. Check Google AI Overviews and AI Mode separately if you have time.

Week 2:

- Check robots.txt and firewall rules for [OAI-SearchBot](https://developers.openai.com/api/docs/bots) and [PerplexityBot](https://docs.perplexity.ai/guides/bots), confirming they can reach you.
- Check indexing of your 20 most valuable pages in Search Console, and look for accidental noindex or nosnippet rules.
- Run a second tracking round.
- Fix factual errors on your own site first: pricing, product names, who you serve.
- Add a free-text question to your demo form asking how the buyer found you.
- Rank your 20 questions by revenue closeness and by how close you are to winning.

Deliverable: a one-page baseline listing the mention rate per engine, top competitors named and your top five target pages.

## What do you do in days 15 to 60?

Publish or improve one page each week, about six in all. Choose improvements first, because existing pages already have links and history.

WeekPageFocus3Core category pageSay what you do, who it fits, who it does not4Pricing pageReal numbers or the factors that set them. Ahrefs saw "price" rise from 6.04% to 8.88% of ChatGPT fan-outs in a month5Comparison against your top rivalFair table, including where they win6Answer page for your highest-value problem questionDirect answer in the first 60 words7Second comparison or alternatives pageBased on who appears in your tracking8Original data or first-hand guideSomething only your company can publish

Use the structure in [how to write an answer page](https://www.fame.so/post/write-answer-page-ai-quotes). The [GEO paper](https://ar5iv.labs.arxiv.org/html/2311.09735) supports adding statistics, quotations and citations, though its main test engine was far simpler than today's products. Get subject experts to review each page, because a specific, correct sentence from a product manager beats three paragraphs of smooth generic copy.

Each week, spend your five hours like this:

- One hour on tracking and the log.
- Three hours on the page.
- Thirty minutes on internal links from related pages.
- Thirty minutes on review and approvals.

Leave the extras for later. Do not spend this period on schema projects or [llms.txt](https://www.fame.so/post/llms-txt-b2b). [Ahrefs' schema study](https://ahrefs.com/blog/schema-ai-citations/) found no clear citation lift among already-cited pages, and the llms.txt evidence is weak too. If either takes under an hour and you have a spare slot, fine.

At day 45, stop and review. Which sources are cited for competitors? Have any fact-check answers improved? Are engines still getting your pricing wrong? Move the remaining weeks toward whatever the log shows, not this table.

## What do you do in days 61 to 90?

Shift effort to where engines already look.

Weeks 9 and 10: map the third-party sources.

- Export every third-party URL cited in your tracked answers.
- Group them into lists, directories, reviews, editorial and community.
- Pick the 15 that appear most and where you honestly fit.
- Claim and correct your directory and review profiles.

Analyze reported a 60.2% mention rate on answers citing a directory page, against 36.3% without, though that comparison is unadjusted. [Ahrefs found](https://ahrefs.com/blog/best-lists-research/) brands higher on third-party lists were more likely to be recommended. It also found 35% of cited lists on low-authority domains, so favor credibility over chasing every one.

Weeks 11 and 12: outreach.

Send five to ten tailored notes a week to list editors. Offer a specific correction or something verifiable, such as a current price or a customer example they can check. Do not offer payment or bulk link swaps. Ask customers for honest reviews where that platform's rules allow.

If you decide to publish your own category list, disclose your role. [Ahrefs' experiment](https://ahrefs.com/blog/self-promotional-content-ai-seo-experiment/) suggests it can help a little-known brand and do little for an established one, and it warns that AI answers may cite the page but recommend someone else.

Week 13: refresh and report.

Update pages with new information, compare results with your baseline, then write the report below.

## What should the day-90 report say?

Keep it to one page:

- Mention rate and citation rate per engine, baseline against the last three-week average.
- Unbranded and branded results, shown separately.
- Competitor mention rates on the same panel.
- Accuracy errors fixed and remaining.
- Pages shipped, and which questions each targets.
- AI referral sessions from GA4, plus demo-form answers mentioning AI tools.
- What you will stop, continue and start next quarter.

Report uncertainty clearly. With 20 questions and three runs per engine, small changes can be noise. [Ahrefs' study](https://ahrefs.com/blog/how-stable-are-ai-citations/) found about 59% of ChatGPT citations gone by the next day, while 75% of dropped citations came back. Model updates can also shift results without any change on your side. Correlation with your work is the most you can claim, and that is fine if you say so.

## What mistakes waste the most time?

- Changing the question panel every week, which removes any baseline.
- Reacting to one missing citation. Ahrefs suggests waiting two to three weeks.
- Treating a citation as a recommendation. They are separate outcomes.
- Publishing many thin pages rather than a few good ones.
- Sending a leader a screenshot of one good answer.
- Letting pricing stay out of date across your own site and directories.
- Ignoring classic SEO. For [Perplexity in particular](https://www.fame.so/post/get-named-in-perplexity), Ahrefs measured high overlap with Google's results.

## When should you get help?

Bring in help if you miss tracking two weeks in a row, if your approval process stops page production, or if engineering holds up crawl fixes. A part-time writer or freelance SEO can take page production while you keep the strategy.

An agency is another option. Compare options by asking for pricing, contract terms, what they ship each month, who owns approvals and what real results they can show. [Fame Search](https://www.fame.so/search) is Fame's own service, listed here for that reason: $2,000 a month, one plan, with weekly tracking, one daily site action, up to 10 new or improved pages a month and outreach drafts, all approved by you. It has no client results yet. Its only evidence is Fame's own site, where 38-76% of tracked AI answers cited fame.so from June 1 to a snapshot on October 30, 2026. Contract terms are not published, so ask on a call. Comparing other providers? See [how Fame compares with FirstIndex](https://www.fame.so/search/vs/firstindex).

## FAQ

**Can one person really do this?**

Yes, if the scope is small. This plan assumes about five hours a week, 20 tracked questions and one page a week. That estimate is a planning assumption, so adjust it to your site and approval process.

**What results should I expect after 90 days?**

No reliable public benchmark exists, so set your own baseline. Expect a clearer view of which questions you can win, fixed technical blocks and more useful pages. Citations may rise, but they fluctuate and cannot be promised.

**Should I stop doing SEO?**

No. Ahrefs found 76% of AI Overview citations also rank in Google's top 10, and Google says standard SEO fundamentals still apply. This plan builds on SEO rather than replacing it.

**What if my leadership wants one number?**

Offer unbranded mention rate, reported per engine as a multi-week trend, and pair it with demo-form answers about how buyers found you. Explain clearly that the sample is small.