GEO strategy: a 5-step, data-driven method

5 steps. Zero guessing.

This is the method we use on our own projects and get results with, shared as is. Every step has an output; none of them rests on a guess.

  1. 01Measure
  2. 02Analyze
  3. 03Set up
  4. 04Research
  5. 05Publish, re-measure

≈ 60 min

Measure

To know what changed, you need to know where you started. Without a baseline, everything you do is a guess.

  1. Sign in to ChatGPT, Claude, Gemini, Perplexity and Copilot (free accounts are fine).
  2. Write the 15–25 questions customers ask before buying: category, comparison, problem and brand questions.
  3. Ask each question in each assistant, in a fresh chat. Are you mentioned, in what position, which competitors, which sources are shown? Log everything in a table.

OutputVisibility rate, average position and citation share: your three-number baseline.

Measurement guide →

≈ 30 min

Analyze

Who do the assistants recommend, and why? The answer usually hides in the sources they show.

  1. Rank competitors by visibility rate in each assistant.
  2. Group the cited sources: list/comparison articles, review sites, forums, guides, brand sites.
  3. List the questions where a competitor appears and you don’t. This gap list is your priority.

OutputGap list + the 5–10 most cited sources + your first 3 actions.

Main guide →

≈ 30 min – 3 h

Set up

AI assistants draw on search indexes. If you are not in the index, you are not in the answer.

  1. Have a site: add it to Google Search Console and submit the sitemap. Add it to Bing Webmaster Tools by importing from Search Console, and turn on IndexNow.
  2. Make sure robots.txt and your Cloudflare settings don’t block AI search crawlers.
  3. No site: your agent builds a simple site and you publish it free on Cloudflare. Connect a domain later, then do the setup above.

OutputVerified Search Console + Bing Webmaster properties, submitted sitemap, open crawler access.

Search Console guide →

≈ 45 min

Research

Don’t guess what people ask; search data already tells you.

  1. Search Console → Performance → Queries: filter question patterns (how, what, best, which, price, vs).
  2. Validate demand with Bing Webmaster Tools → Keyword Research and the AI Performance grounding queries.
  3. Expand the list with Google autocomplete, “People also ask” and Google Trends.

OutputA prioritized question list with evidence for each.

Keyword guide →

every 2–4 weeks

Publish, re-measure

One clear answer page per priority question, then a re-measure with the same question set. That is the loop.

  1. Answer in the first 2–3 sentences. Add numbers, quotes and source links. Headings, a table, an FAQ.
  2. Request indexing in Search Console and notify Bing with IndexNow.
  3. Honestly earn a place on the third-party pages the assistants cite.
  4. Run the same question set again, compare against the baseline, report.

OutputA before/after report and the next 3 actions.

Bing + IndexNow guide →

Shortcut

Let your agent run these 5 steps.

The agent asks you questions first, then measures, sets up and reports every 2–4 weeks. It never asks for passwords and asks before publishing anything.

Paste into your agent

Read the GEO Asked method at https://geoasked.com/agent.md and run it for my brand, step by step. Ask me short questions first.

ajan@geoasked:~$ curl /llms.txt
Loading…
Agents start with these files →agent.mdllms.txtllms-full.txt