How it works

Three LLMs. One clear verdict.

How KaptainLLM produces a scan — from the moment you submit a URL to the PDF landing in your inbox.

Step 01

You submit a domain, a competitor, and 3 customer questions

Your domain, one competitor domain, and three actual phrases your buyers type into AI. Plus a market (UAE, KSA, Egypt, or global) and a language (English or Arabic). Nothing else required for the free tier.

Step 02

We fan out 9 queries across 3 LLMs

For every question, we run it against three models: OpenAI's GPT-5, Google's Gemini 2.5 Pro, and Perplexity's Sonar Pro. Each gets market framing embedded — so the LLM answers as if the user is physically in your market. All queries run with live web search on.

Step 03

We score every query: won, shared, or lost

Won — your brand/domain was cited by at least one of the 3 providers. Shared — you AND your competitor both appeared. Lost — only your competitor was cited. Neither — the AI didn't name either of you.

Step 04

Claude analyzes the raw responses and builds your verdict

Anthropic's Claude reads all 9 LLM responses verbatim and generates a one-line verdict, a diagnosis of why the gap exists, and three ranked actions you can take this week. Not generic advice — specific to what the LLMs actually said.

Step 05

Report rendered and emailed

Rendered to a clean A4 PDF — verdict, scoreboard, per-query responses with citations, diagnosis, actions. Sent to your inbox as an attachment, plus a web version linked from your dashboard.

Fine print

The details we don't hide.

Model versions are fixed

GPT-5, Gemini 2.5 Pro, Perplexity Sonar Pro. We don't change these without notice — your Monday scan and your Friday scan are comparable.

Responses vary; that's the point

LLM responses are probabilistic. Running the same query twice can give different results. We capture one snapshot per scan. Track month-over-month to see the real signal.

Arabic is native, not translated

When you pick Arabic, we send the query in Arabic with market framing ("أنا في الرياض..."). Responses come back in Arabic. Claude then reads the Arabic responses directly to build your insights.

We store every scan

Your scans stay in your dashboard. Re-run them next month to compare. Export history anytime.

Ready to see yours?