Your docs get cited. Your competitor gets recommended.

What is AEO for developer tools?

Short answer

Making sure AI assistants recommend your tool when a developer asks for one – and describe it correctly when they do.

AEO, GEO, LLM SEO, AI search optimization – four names for one job. Developers ask ChatGPT, Claude, Perplexity or the assistant inside their editor "what's the best X for Y", get three or four names back, and go straight to those tools' docs.

SEO gets your pages ranked. AEO gets your product into the answer. They overlap – models still read the web – but the pages that win are different. A model lifts verdicts, tables and explicit "best for" statements. It rarely lifts your 3,000-word explainer.

Why is your organic traffic dropping?

Short answer

Because the research step moved into the AI answer. Your site now sees buyers late, when they're already comparing.

The buyer journey starts in ChatGPT, not Google. The visits you lost were mostly early-stage reading – the "what is" and "how to" traffic that rarely converted anyway. The visitors who still arrive are further along and more decided.

So falling traffic on its own isn't the problem. Not being on the shortlist is. A developer who never hears your name in the answer doesn't visit, doesn't search for you later and doesn't show up in any dashboard you own.

Three ways a devtool goes invisible in AI search

Short answer

Diagnose which one you have before you build anything – each needs the opposite fix.

1. Cited, not recommended

   Your "what is vector search" post is the source the model quotes – and then it recommends a competitor. More educational content makes this worse. The fix: comparison and "best for" pages that give the model a verdict to repeat.

2. Absent everywhere

   The sources the model trusts in your category – listings, comparison articles, community threads, GitHub – don't mention you. Your own site can't fix this alone. The fix: get into the third-party sources already cited for your prompts.

3. Described wrong

   Old pricing, the wrong use case, "not production-ready" from a two-year-old thread. Worse than being absent: the buyer walks away with a false picture and never checks again. The fix: your own pages first, then the high-authority listings repeating the error.

Most teams have a mix, and it differs by prompt: absent for "alternatives to [big incumbent]", described wrong for "[your tool] pricing", cited-not-recommended for the category question.

Does llms.txt or schema fix it?

Short answer

No. They tell you whether a model can read you, not whether the sources it trusts mention you.

llms.txt and schema are cheap, so add them. But some of the most-cited tools in their categories don't have an llms.txt at all. If a vendor's audit ends with "add llms.txt and FAQ schema", it has checked whether your site is machine-readable – not why you're missing from the answer.

What does the AI visibility audit cover?

Short answer

Four weeks, around 50 buying-intent prompts, and a roadmap where every fix has a success metric and a kill switch.

1. Week 1: Build the prompt set.

   About 50 prompts your buyers ask, written with you: "best X for Y", "X vs Y", "alternatives to Z", "how to migrate from Z". Not 1,000 broad ones – those produce a nice-looking score and nothing you can act on.

2. Week 2: Run and read.

   ChatGPT, Claude, Perplexity and Google AI Overviews, with repeated runs because answers change from one run to the next. Who gets recommended, which pages get cited, and how you're described.

3. Week 3: Diagnose.

   Which of the three shapes you're in, prompt cluster by prompt cluster. Which sources the models trust in your category, and what your competitors have there that you don't.

4. Week 4: Roadmap.

   A prioritized list of fixes, each with what success looks like and when to kill it. The baseline you'll measure against. If you're hiring an SEO or AEO lead, the scorecard for that hire.

You keep the prompt set, the baseline, the source map and the roadmap – rerun it yourself any time.

What happens after the audit?

Short answer

Your team runs the roadmap – or I build the first fixes with you on a monthly retainer.

If you take the retainer, I build what the audit says to build, in order:

• Granular comparison pages. Not one "us vs everyone" page. Three competitors × three angles – pricing, features, reviews – each with a verdict, a table and explicit "best for" statements. Those are the parts models lift.
• Corrections. Your own pages first, then the listings and directories repeating the wrong description.
• Third-party presence. The comparison articles, community threads and listings already cited for your prompts.
• Monthly measurement. The same prompt set every month, plus an open-text "how did you hear about us?" field on signup.

Last-touch attribution will credit Google or direct for buyers who found you in ChatGPT – and you'll cut the channel that's working.

What does this look like in practice?

Short answer

Granular comparison pages turned ChatGPT into a top-five acquisition channel in about three months.

At a developer-tools company where I ran marketing, we built separate comparison pages against each of the main competitors – pricing, features and reviews, each page with a verdict and a table. They ranked in Google and started appearing in AI answers. Within about three months, ChatGPT was a top-five acquisition channel.

AI visibility audit or a full-time AEO hire?

Short answer

It isn't either/or. The audit equips whoever you hire.

Hire next month or next year – the work still counts. The roadmap and the candidate scorecard are what a new hire would otherwise spend their first quarter producing.

Who is this for – and who isn't it for?

Short answer

Developer-facing B2B products with a content function already in place and a traffic problem you've already noticed.

A fit if:

• You sell a developer-facing B2B product – API, infrastructure, database, dev platform – roughly Series A–B.
• A content or SEO function already exists, and traffic or CTR has been dropping for months.
• You've decided someone needs to own this, and you're weighing a hire.

Not a fit if:

• You're pre-seed with no content yet – you need something worth recommending first.
• Your team already runs AEO in-house with its own measurement.
• You want a free audit or a guaranteed spot in ChatGPT. Nobody can promise the second, and anyone who does is selling the first.

Who you'd be working with

I'm Ilia Markov. At Meilisearch I built the marketing function from zero for a technical product sold to developers – including comparison pages built for AI search before anyone called it AEO. As Marketing Director at Toggl, I rebuilt the product-led engine. Before that, content and growth at ChartMogul, Hubstaff and Groove.

I write the prompt sets, read the answers and build the pages myself.

MBA from Michigan Ross. Based in Sofia, working remotely since 2015.

Questions devtools founders ask

Ask ChatGPT for the best tool in your category. Don't like the answer? That's the first call.

30 minutes. Bring three prompts your buyers ask – you'll leave knowing which shape of invisibility you're likely dealing with, whether or not we work together.