AI search runs on skills

A new surface,

same fundamentals.

Buyers now also ask AI engines, and those engines answer by retrieving sources and citing them. That makes this a second index with its own metrics - and one that rewards the fundamentals you already run. The organic channel did not change hands; it grew another surface.

Three questions decide whether you show up there, and you already know how to work all three: can the engine find your pages, can it understand them, and does it cite them. Crawlability, structured data, internal linking and earned authority are what move all three - read by a different consumer. Attention is moving with it: Gartner projects a 50%+ shift away from brands' organic search traffic by 2028 as consumers adopt generative search. Same buyers, asking somewhere else.

The log file

Google Analytics cannot see.

You already know how to read a log file. Here is the one your analytics stack never shows you: GPTBot, ClaudeBot, PerplexityBot and the rest send direct HTTP requests and never execute JavaScript, so GA4 records nothing at all. BotsGenie reads them straight from your CDN logs - Cloudflare, Vercel, Fastly, roughly a five-minute Worker setup - and classifies every visit by intent.

The Overview then does the part that survives scrutiny: it overlays GA4 so bot crawls, citations, AI-referred sessions and revenue sit in one funnel, per engine. We have analysed 2.7 million crawler-log rows of this traffic - what that data actually shows.

Technical GEO

is your home turf.

SiteGenie audits the things you already argue for in sprint planning: structured data built for LLM indexing, AI crawlability, llms.txt, internal links and entity clarity. In our own scans 35% of pages shipped with no structured data at all, and connecting entities into one @graph rather than isolated snippets took a site from 13 to 123 connected nodes with 24 of 25 types validating. What schema really does for AI search covers the honest version of that argument.

Measurement that

survives a QBR.

Three numbers, defined tightly enough to defend in front of a CFO:

• Visibility % - how often your brand appears at all for the prompts you track.

• Share of Voice % - how prominent you are when you do appear, against your named competitors.

• Citation Rate % - how often an engine cites one of your pages as the source.

Read together they produce the diagnostic you will recognise instantly: the Visibility-Citation Gap - engines mention your brand but will not use your page as a source. It is the AI-search analogue of ranking without earning the click, and it has a different fix than simply being invisible.

Keep the per-engine split visible, because averages hide the story. In our own July tracking the same brand with the same content scored 6.8% visibility on Gemini and 0.0% in Google AI Overviews. How to read the three metrics together.

From insight

to a shipped fix.

Recommendations become editable briefs with the real citations pulled in, and the Article Generator turns an approved brief into a full draft in your brand voice, governed by Content Guidelines you set once. For the gap that is not on your domain, the Publisher Network places content on sites the engines already cite. The deep version of that workflow lives on ActionsGenie.

Fits the stack

you already run.

An official read-only MCP server lets you query your own visibility data from Claude, ChatGPT or Cursor, so the reporting you want exists whether or not we built the view. CSV export, CDN connections for Cloudflare, Vercel and Fastly, and coverage across the six engines sit alongside it - including index-presence checks, because Bing feeds both Copilot and ChatGPT's web answers and "indexed in Google, missing in Bing" is a real and fixable gap. Why one index is not enough.

Frequently asked questions