A new layer of business intelligence
Measure what the models measure.
Combot provides the authoritative, non-obvious metrics for how large language models see your brand — from training-data memory to final recommendation.
Request access See what we track →01 — The new funnel
Beyond mentions.
Your brand's visibility inside an AI answer is not a binary "yes/no." It's a 7-layer funnel from a model's latent knowledge to its final, commercially-weighted recommendation. Combot's AI Share of Recommendation is the north-star metric for the AI era.
Read the full framework in The 7 layers of AI visibility.
02 — Ground truth
Your data, modelled.
Combot integrates with the systems you already run and consolidates them into a single BigQuery-backed data lake. The Claude tool-use loop reasons across every source at once — not just one channel, not just one chart.
Combot's own reasoning runs on a multi-model orchestra — Anthropic Claude, OpenAI, and Google Gemini — but those are the engine, not your integrations. Your data stays in your warehouse.
03 — Where you already work
Chat-native. Where your team already works.
Combot lives in your team's chat — Slack, Teams, Mattermost, or any platform you connect via API integration. Trigger a full SEO audit, ask for a weekly anomaly report, query the data lake, or pull live SERP — without leaving the channel. No new dashboard to learn. The answer comes back where the conversation already is.
Conversational command
Ask Combot in plain English. It chooses tools, runs queries against BigQuery, calls live APIs, assembles the answer. No SQL needed for the common cases — and full SQL passthrough for the rest.
Anomalies surfaced, not searched
Multi-channel correlation: when a spend spike on Ads doesn't match a click pattern in GSC and CWV regressed at the same time on the same template, Combot flags the compound — not three siloed warnings.
Tenant isolation by design
Each client lives in its own GCP project with its own BigQuery datasets, its own service account, its own audit trail. Multi-tenant where it counts; isolated where it must be.
04 — From the blog
Notes on building the AI visibility stack.
AI Share of Recommendation: a new north-star metric
Why "are we mentioned in ChatGPT?" is the wrong question, and what to track instead.
The 7 layers of AI visibility
From training-data memory to agent action. Each layer fails differently — and each requires different work.
Source mapping: tracing an LLM's answer back to its roots
The citation is not the source. The source is whoever wrote the page the model paraphrased from.
Fact-checking the machines: building an automated accuracy center
Hallucination isn't a bug to dunk on — it's a KPI to track, alert on, and remediate.
From fetchability to trust: the technical SEO of language models
AI bots crawl differently. Most don't execute JavaScript. Most never reach your category pages.