Designing Website Tools for AI Agents with WebMCP
WebMCP is a September 2026 draft for exposing website functions to browser agents as structured tools.
Read insight ->News
Insights for brands being evaluated by AI.
A practical editorial hub for teams that need to understand how English AI search changes discovery, comparison, and trust.
When an AI Answer Should Stop at the Evidence Boundary
An AI system should withhold a conclusion when the evidence available for the task cannot support it.
Read insight ->The Context Budget and the Evidence an AI Answer Leaves Behind
A retrieved page can fail to influence an AI answer because its relevant passage was never passed to the generator, was compressed without its qualification, or was available but not used correctly.
Read insight ->Managing Google Search AI Inclusion at Page and Folder Level
Google now documents a Search Console control for inclusion in specified generative AI features on Google Search.
Read insight ->Reading Google AI Impressions Without Double Counting Exposure
Google's generative AI performance report measures impressions from supported Search AI features, but its chart and page table do not always use the same aggregation.
Read insight ->Writing Useful FAQ Content After Google Retired FAQ Rich Results
FAQ content can still be useful after Google retired its FAQ rich result, but the value now has to come from the information itself.
Read insight ->Counting Evidence Without Counting the Same Source Twice
Several websites repeating a claim do not necessarily provide several independent reasons to believe it.
Read insight ->Calibrating AI Judges Before You Trust a GEO Quality Score
An AI judge can help review large numbers of answers, but its score is a measurement instrument that needs calibration.
Read insight ->When Retrieved Text Tries to Give Instructions
Retrieved content should be treated as evidence to inspect, not authority that can redefine an AI system's task.
Read insight ->The Product Variant Problem: Matching the Right Item to the Right Question
AI shopping answers become unreliable when a product family is treated as one undifferentiated item.
Read insight ->Designing a Prompt Panel That Survives Review
A defensible GEO prompt panel is a versioned sampling instrument, not a list of convenient questions.
Read insight ->The PDF Retrieval Problem: When a Readable Document Is Hard to Extract
A PDF can be perfectly readable to a person and still be unreliable evidence for an AI system. The visible page is only one representation.
Read insight ->
12 published insights on this page.