We asked 5 models 'best CRM for a small team' 300 times. Here is what happened.
Aug 2026 · 9 min read
Resources
Everything we know about GEO, written down.
The GEO Guide
Generative Engine Optimization is the practice of making sure large language models cite, recommend and accurately represent your brand in AI-generated answers. This is the short version.
- 01Why AI answers are a distribution channel4 min
- 02The dimensions of GEO7 min
- 03Machine access: robots.txt, llms.txt, rendering6 min
- 04Answerability and structured content8 min
- 05Entity consistency and the knowledge graph5 min
- 06Citations, reviews and third-party authority6 min
- 07Measuring share of voice against competitors5 min
- 08Building a 90-day GEO roadmap6 min
Blog
Field notes on AI search
Your pricing page is probably invisible to ChatGPT. Here is the fix.
Aug 2026 · 5 min read
llms.txt, explained: what to put in it and why it matters
Jul 2026 · 6 min read
Share of voice is the only GEO metric your CEO will remember
Jul 2026 · 4 min read
Case studies
Before and after
Northwind
Analytics
54→7192 days
“From invisible to shortlisted on every model.”
Halcyon Labs
DevTools
47→6988 days
“Prompt evidence got engineering to prioritise it.”
Cedar & Vine
DTC
61→7890 days
“Doubled our AI-assisted revenue in a quarter.”
FAQ
Common questions
Weighted dimensions — AI Visibility, Brand Prominence, Recommendation Performance, Competitive Position, Citation Visibility, Authority & Trust, Technical GEO and Entity Consistency — each scored 0–100 and calibrated against a benchmark set. The exact set of dimensions comes from the audit backend.
ChatGPT (GPT-4o), Claude, Perplexity, Google AI Overviews / Gemini and Microsoft Copilot, depending on which providers the audit backend is configured to run. Each prompt is run multiple times per model and aggregated.
You can run a free audit and see your score without signing up. An account is used to save audits and track progress over time.
SEO optimises for ranked links on a results page. GEO optimises for being cited, recommended and represented accurately inside an AI-generated answer, where there is often no list of links at all.
Typically 1–4 minutes. The processing screen shows live status while the backend crawls your site and probes the configured AI engines.
Yes. Re-running produces a fresh score against the current state of your site and the models.
Documentation
How the audit works
phazeAi measures how large language models and AI answer engines perceive, retrieve and represent a brand. It combines live model probing, citation extraction and a technical crawl.
Prompts
Buyer-intent prompts derived from keyword research and interviews
Models
The AI engines the audit backend is configured to run
Runs
Each prompt run multiple times per model and aggregated
Crawl
Your site fetched both rendered and raw-HTML
Scoring
Weighted dimensions, 0–100, calibrated against a benchmark set