Buyers ask ChatGPT, Claude and Google AI Overviews for a development partner and get a list of five names. We get your company into the lists those answers come from. Then we measure how often you are named, before and after.
"best .net development companies"
"top fintech software development companies for a mid-size bank"
"we need a logistics app built, which outsourcing company would you recommend"
Each prompt returns three to seven names. If you are not on the list, the buyer never visits your site. Buyers use four kinds of prompts: problem-aware ("we need a logistics app built"), vendor-discovery ("best .NET development companies"), comparison ("X vs Y") and trust ("is X reliable"). The campaign targets vendor-discovery and comparison prompts; your own site answers the other two.
Not a theory. The searches below are what ChatGPT and Claude actually ran when we asked them for a development partner.
The assistant does not look up your name. It splits "best .NET development companies" into searches: the category and its synonyms, the stack, the vertical, "2026", "case studies". In our tests ChatGPT ran 3 to 7 searches per answer, Claude 2. Google calls this query fan-out.
Directory pages, media roundups, review sites and vendors' own lists. ChatGPT fetches them live with its crawlers; Google AI Overviews read Google's index. If you are in none of these pages, you are out before the model has formed an opinion.
The model keeps the names that several sources agree on and drops the rest. One list is an opinion. Five lists with the same five names is a consensus, and the "best for" lines are copied from those pages.
Before recommending, the model checks the candidate's own site and profiles: does the service page say the same, do Clutch and GoodFirms agree, are OAI-SearchBot, ClaudeBot and Googlebot allowed in robots.txt. A mismatch or a blocked crawler, and the name is dropped.
Our work is step 2: we put you into the lists those searches return. The lift in visibility shows one to two weeks after the campaign.
Classic SEO meant ranking a page on Google. LLM SEO means being named when a buyer asks an AI for a development partner. It is also called generative engine optimization (GEO), answer engine optimization (AEO) or AI search optimization. Same job, different source: the answer is built from lists, not from your page.
AI answers about service companies come from lists: directory pages, media roundups and vendors' own lists. A Clutch profile makes you a source. It does not make you a recommendation.
A recommendation needs several independent lists that agree on you. Five lists with the same five names is what the model repeats.
Your own blog helps little. Third-party lists get most citations, self-promotional ones about a tenth.
Peec AI, analysis of 232,000 citations, December 2025 to February 2026.
We add the missing lists: on sites the crawlers read, naming you next to the competitors you lose to today.
We start with a narrow question. "Best software development company" is not for sale here.
Your stack, vertical and geography. "Ruby on Rails development companies in the Netherlands", not "best software development company".
10 answers per prompt from ChatGPT, Claude and Google AI Overviews before anything is published.
30 listicles, published in stages over 10 working days. Each names you next to real competitors, on sites LLM crawlers visit. We pick the sites by crawler data, not domain rating.
Daily tracking. Report in week 3 to 4: mention share per surface, before and after, with dates.
Measured on live campaigns, with dates. Your category may move slower. We say so before you pay.
Campaign 1: "best .NET development companies" and its variants. 30 listicles live by 21 July 2026. 120,861 AI crawler visits to the placements. Tracking since 23 June, report 10 August 2026.
| Share of checks naming Leobit | Before | After |
|---|---|---|
| Google AI Overviews | 4% | 36% |
| ChatGPT | 12% | 52% |
| Claude | 2% | 66% |
“We have already received three relevant leads from ChatGPT and Claude.”
Presence in Google AI Overviews for its category after the campaign.
Leobit is now running its second campaign with us.
Enough to build a consensus where few names compete. Most dev shops start here.
Twice the volume for a category where established vendors already sit in the lists.
Scoped on the call.
Tracking for a month is included. The placement report lists every live URL.
We show you the prompts your buyers run, who the models name today, and the narrow question where 30 placements are enough. Free, whether you buy or not.