A 30-day AEO plan for a team of one

Four weeks, an hour or two a week, in dependency order: access first, then identity, then evidence, then measurement.

Most AEO advice assumes a team. This assumes you, an hour or two a week, and a list of forty other things that are also on fire.

The order matters more than the pace. Each week gates the next: there is no point earning citations for a site the crawler cannot fetch, and no point writing an About page for a site with no coherent name. If you fall behind, stay in order rather than skipping ahead.

Week 1 — Access

Goal: a crawler can fetch your pages and get real HTML. This is binary, it is cheap, and it invalidates everything downstream if it is broken.

• Read yoursite.example/robots.txt. Find any AI user-agent group you did not deliberately create. What each one is.
• Decide, once, on training crawlers versus answer crawlers. Write the decision in a comment in the file. The trade-offs.
• curl -sI -A "...GPTBot/1.2..." https://yoursite.example and check for a 200. A 403 means your WAF is overriding your intent. Fixing that.
• curl -sL https://yoursite.example | head -100. If it is an empty <div id="root">, you have a rendering problem and it is now the most important thing on this list. Why, and the minimum fix.

Done when: a bot user-agent gets a 200 with real markup in the body.

Week 2 — Identity

Goal: a machine can say what your company is called and what it does.

• Write down your canonical set: brand name, legal name, canonical URL, one-sentence description, logo URL. This is a decision, not a task, and it is the hour that matters most. The five fields.
• Add or fix your Organization JSON-LD on the homepage, server-rendered. Field by field.
• Check you picked the right @type. If customers physically visit you, it is LocalBusiness. The rule.
• Grep your templates for stale variants of the brand name — footer, OG tags, email templates, site title config.
• Rewrite the first paragraph of your About page so it states who, what and for whom in three sentences. What good looks like.

Done when: an assistant asked "what is <your brand>?" in a fresh session gives a correct, unhedged answer.

Week 3 — Evidence

Goal: sources you do not own agree with you.

• List every profile your company has. Kill or claim the dead ones.
• Update the survivors so name, description, logo and website match the canonical set exactly.
• Claim and complete your Crunchbase profile. Check the category is right. Why it is first.
• Claim your review-site profiles if your category has them, and fix the category selection — it is frequently wrong and it decides which comparisons you appear in.
• Add every profile URL to your schema's sameAs. The connective tissue.
• Twenty minutes: write /llms.txt. Last, because it is the smallest of these. What it is and is not.

Done when: sameAs lists four or more live profiles that all describe you the same way and all link back.

Week 4 — Measurement and the next quarter

Goal: know where you actually stand, and have a list for next quarter.

• Write five buyer questions with no brand name in them. Save them — you will reuse this exact set every quarter.
• Run them. Record the hit rate, the competitors named, and the sources cited. How to read the three numbers.
• Take the top three competitors and do the ten-minute entity read on each. The six checks.
• Take the cited sources and sort them: present and correct, present and wrong, absent. Fix the wrong ones first — that is the fastest win on the list. Source by source.
• Pick two absent sources to pursue over the next quarter. Two, not ten.

Done when: you have a number, a competitor list, and two named targets.

What month two looks like

Nothing on this list repeats weekly, which is the point. After the first month the work is:

• Re-run the five questions once a quarter, not weekly. The noise floor is higher than any change one week produces.
• Re-scan after any site change that touches templates, hosting, or your CDN configuration — those are how access breaks silently.
• Keep pursuing the two off-site targets. This is the part that takes months and it is the part that compounds.

The realistic expectation

Weeks 1 and 2 are the ones with a fast feedback loop — fix an access problem and it can show up within weeks. Week 3 compounds slowly. Week 4 tells you which of the other three to spend more on.

And the honest ceiling: if a competitor is named because they are genuinely the better fit for the question, no amount of markup changes that. This plan removes the reasons a machine has to leave you out. It does not manufacture reasons to put you in.