AI Search
How often does ChatGPT search? The 14.4% figure needs a correction.
Ahrefs corrected its headline metric from web searches to recorded fan-out queries. Here is what that changes for AI visibility measurement.
Understanding AI discovery
What AI search actually does.
How information is discovered, selected, trusted, and cited by AI systems.
The lead story · Retrieval
How often does ChatGPT search? The 14.4% figure needs a correction.
Ahrefs corrected its headline metric from web searches to recorded fan-out queries. Here is what that changes for AI visibility measurement.
Read the analysis ↗A fan-out query, a web search, and an answer are different units. A useful measurement starts by telling them apart.
Source-led analysis · 4 min readLatest coverage
Seven questions worth asking
Measurement
The AI attribution gap does not come with a universal multiplier
The often-repeated 10x figure traces to one company’s survey comparison. Here is a better way to reconcile AI referrals, discovery, and revenue.
Technology
Read the counts behind the crawl-to-refer ratio
A lower ratio can mean less crawling without a single extra visitor. Separate bot purpose, delivery cost, and referral outcomes before changing access.
Technology
When AI agents fail, inspect the interface
Missing labels and misleading controls can break agent journeys. Test the repair on a real task before assigning an agent-readiness score.
GEO
Does schema markup reach AI answers? Separate the stages.
Google requires no special schema for AI search. The useful question is where a fact survives: on the page, in a search index, in retrieved text, or in the final answer.
Measurement
What “How did you hear about us?” can measure about AI
Customer surveys can reveal remembered AI influence. They cannot, on their own, recover a complete journey or prove incremental revenue. Better questions and denominators make them useful.
Measurement
Before explaining an SEO swing, audit the measurement
Rank snapshots, analytics sessions, and business outcomes can disagree for different reasons. A measurement audit separates collection faults, missing attribution, and genuine performance changes.
Explore the disciplines
Six editorial desksNews
Breaking developments, model releases, platform changes, industry announcements, and acquisitions.
AI Search
How answer engines retrieve, synthesize, and present information.
GEO
Generative engine optimization, entity authority, citations, content optimization, information gain, and brand discoverability.
Research
Academic papers, patents, retrieval experiments, benchmarks, original studies, and emerging theories.
Measurement
AI visibility, prompt tracking, citation attribution, share of voice, referral analytics, and measurement methods.
Technology
RAG, embeddings, vector databases, knowledge graphs, agents, MCP, crawling, indexing, and retrieval infrastructure.
From evidence to action
One finding.
Three perspectives.
Research
Establish what happens through studies and experiments.
Measurement
Determine how to quantify it and what the numbers support.
GEO
Translate the evidence into optimization decisions.
Our approach
Show the source.
Check the claim.
01
Go to the original
Follow figures back to the study, documentation, or dataset that produced them.
02
Keep the conditions
Preserve the sample, time period, model, and denominator. They determine what a finding means.
03
Make it useful
Turn the evidence into a measurement decision, a practical check, or a better question.