GEO
What belongs on a GEO checklist for a developer tool?
Updated October 2026 · Manoj Surya
Quick answer
A GEO checklist for developer tools covers four things: pages that answer buyer questions in the first sentence, accurate docs and pricing, useful presence in the communities developers trust such as GitHub, Hacker News and Stack Overflow, and a fixed question list re-checked weekly. MarketHQ monitors those communities on every plan and runs the weekly AI-answer check on paid plans.
Key points
- Start with the question list; every other item on the checklist is judged against it.
- Fix your own pages before chasing mentions: task docs, comparison pages and a pricing page written as text.
- MarketHQ monitors Hacker News, GitHub, Stack Overflow, Dev.to and Lobsters on every plan, so you can find threads worth answering.
- MarketHQ re-asks your question list in six AI engines every week on paid plans and shows a four-week trend.
Founder of MarketHQ. Building market intelligence for companies that sell to developers.
This is a working checklist for a developer-tool company that wants AI assistants to name it when developers ask for a tool. Each item says what to do, how to tell it is done, and how often to revisit it. No single item guarantees a mention; together they make your product easier to find and to describe correctly.
What is on the GEO checklist?
Ten items in three groups: the questions you track, the pages you own, and your presence where developers talk.
Work through it roughly in order. You control items 1 to 7 fully; items 8 to 10 depend on other people and the engines, so they need a regular habit.
| Item | Done when | Revisit |
|---|---|---|
| 1. Question list | You have 10 to 20 written questions covering tasks, tool choice, alternatives and migration | Every quarter |
| 2. Task docs | Each common 'how do I X with Y' question has a docs page that opens with the answer and a working sample | When the API changes |
| 3. Comparison pages | A '<your tool> vs <competitor>' page exists for each competitor that shows up in your answers, with a dated table | Every quarter and when pricing changes |
| 4. Alternatives page | A page answers 'alternatives to <competitor>' and names that competitor plainly | Every quarter |
| 5. Pricing page | Plans, limits and prices appear as text in the page HTML | Every price change |
| 6. README | The first paragraph says what the project does, who it is for, and the install command | Each major release |
| 7. Changelog | Dated entries on a public page that loads without scripts | Every release |
| 8. llms.txt (optional) | A short markdown file at /llms.txt links to your key docs pages | When the docs structure changes |
| 9. Community presence | You answer relevant threads on Hacker News, GitHub, Stack Overflow and Reddit, with affiliation disclosed | Daily or weekly triage |
| 10. Weekly re-check | Same questions, same engines, same day, with who was named and which links were cited logged | Every week |
How do you build the question list for a developer tool?
Collect the exact wording developers use in issues, threads and search queries, then group it by question type.
Developers attach a framework or language to their questions, as in 'how do I stream responses from a FastAPI endpoint'. Keep that wording.
- GitHub: search issues on your repo and your competitors' with qualifiers such as is:issue and in:title to find recurring 'how do I' and 'does it support' questions.
- Stack Overflow: read recent questions in your tag and your competitors' tags.
- Hacker News: search Ask HN threads on hn.algolia.com for your category.
- Google Search Console: filter your queries for question words and 'vs' or 'alternative'.
- Sales calls: copy prospects' questions word for word.
Which page fixes come first?
Start with the question where a competitor is named and you are not, and fix the page that should answer it.
Task docs and comparison pages usually come first because they map directly to questions. The pricing page is next. An assistant asked 'how much does <your tool> cost' can only repeat what it can read, and a price shown only inside an image, a calculator widget or a 'contact us' button gives it nothing current to use.
llms.txt is a proposed convention: a markdown file at your site root that lists the pages most useful to language models. It is cheap to add. How much, if at all, the major AI engines use it is unclear, so treat it as optional and do it after the pages above are in good shape.
How do you show up in developer communities without spamming?
Answer the question that was asked, say that you work on the tool, and mention it only when it actually fits.
Hacker News, GitHub, Stack Overflow and Reddit each have their own norms. A Show HN post works for a launch. Stack Overflow expects you to disclose affiliation when you recommend your own product. Many subreddits limit self-promotion, so read the rules before replying. Across all of them, a reply that solves the problem first and names your tool second is the kind people upvote and link to later.
MarketHQ finds the threads. Every plan, from Lite at $29 a month, monitors Hacker News, GitHub, Stack Overflow, Dev.to, Lobsters and blogs, checks four times a day and picks up to three reply opportunities, plus Reddit, X, LinkedIn, YouTube and Bluesky, and drafts replies you review and post yourself. MarketHQ never posts for you.
How do you run the weekly GEO re-check?
Ask the same questions in the same engines on the same day each week and record the same fields every time.
Log, for every answer: the engine, whether you were named, which competitors were named, the links cited, and the full answer text. Compare four weeks before deciding a fix worked, because answers vary between runs.
MarketHQ's paid plans do this every week starting Monday, in ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, up to 10 questions on Lite, 60 on Startup and 120 on Growth. The AI Visibility page shows the latest answers, a four-week trend line and the brands AI recommends most in your space. Weekly gap analysis turns the questions you lose into an action plan with evidence attached, which feeds back into items 2 to 9.
What does a first month on the checklist look like?
Illustrative example: a database migration tool works through the checklist in four weeks and sees a small, early change.
Illustrative example: in week one, a database migration tool writes 12 questions and records a baseline: it is named in a few ChatGPT answers and none from Claude. In week two it moves pricing into the page text and publishes a comparison page against the competitor that appears in most answers. In week three it answers three Stack Overflow questions in its tag, with affiliation disclosed, and replies in one Ask HN thread.
By week four it is named for one more question, and the Claude answers still carry the 'Answered from memory, no sources' label. A month is too short to call a trend, so the team keeps checking the same questions.
What will this checklist not do?
It improves your odds of being named; it cannot guarantee a mention or control what a model learned in training.
Answers vary by user, account and run, and MarketHQ asks through each engine's API, so a signed-in developer may see something else. MarketHQ does not track Google AI Overviews, Google AI Mode or Microsoft Copilot, and it does not audit your site, generate schema or write llms.txt for you.
The checklist is written for developer tools, but the structure works for any company: swap Hacker News and Stack Overflow for the forums and review sites your buyers use.
FAQ
- What is the first item on a GEO checklist?
- The question list. Write 10 to 20 questions developers ask, in their own words, before they know your name. Every page fix and community reply on the checklist is then judged by whether it helps you get named for those questions.
- Does llms.txt help AI engines find my docs?
- Possibly, but nobody outside the engines can say for sure. llms.txt is a proposed convention, and how much the major AI engines use it is unclear. It takes little time to add, so do it after your task docs, comparison pages and pricing page are in good shape.
- How often should a developer tool re-check AI answers?
- Weekly, on the same day, with the same questions. Daily checks mostly pick up the normal variation between answers. Look at four weeks together before deciding a change worked.
- What tool runs the weekly re-check on this checklist?
- MarketHQ: paid plans ask your tracked questions in ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek every week, log who was named and which links each answer cited, and turn lost questions into an action plan with the evidence attached.
- Is there a free way to start this checklist?
- MarketHQ has no free plan for the community items; Lite, from $29 a month, checks Hacker News, GitHub, Stack Overflow, Dev.to, Lobsters and blogs four times a day for your brand and two competitors, with a 4-day trial. For a free, no-account read today, use the free GEO audit: a one-off reading in ChatGPT and DeepSeek.
Related
- Free GEO audit — one-off reading in ChatGPT and DeepSeek
- Track mentions on Hacker News, GitHub and Reddit — finding the threads on the checklist
- Build an AI visibility prompt list — item 1 in detail
- Monitor GitHub issues for competitors — mining issues for real questions
- AEO for developer-tool companies — docs, README and Stack Overflow in depth