Practical research on AI discovery
Flagship research
A practical way for a small team to make useful evidence easier for people and answer systems to find, assess, and cite—without treating citations as a promise.
Read the researchFrom the desk
guides
Turn Bing-native citation data into a useful page review. Preserve filter states, inspect query–page mappings, and interpret Citation Share without mistaking sampled reporting for a rank, a buyer prompt, or a competitor list.
guides
A redirect can work while the supporting claim disappears. Before changing domains, CMS, or URLs, map the evidence buyers need, review each successor, and separate release-quality checks from later AI citation observations.
guides
Your website may be current while an old report still gives buyers last year's numbers. Audit the downloadable evidence alongside its landing page: check the text, dates, claims, public URL, and revision history before using a citation observation to guide AEO work.
guides
An English prompt run from your office cannot tell you how a German-speaking buyer in Austria researches your category. Build separate observations for language, market, and buying task, then audit whether the public evidence actually supports the offer in that market.
Research section
guides
Turn Bing-native citation data into a useful page review. Preserve filter states, inspect query–page mappings, and interpret Citation Share without mistaking sampled reporting for a rank, a buyer prompt, or a competitor list.
guides
A redirect can work while the supporting claim disappears. Before changing domains, CMS, or URLs, map the evidence buyers need, review each successor, and separate release-quality checks from later AI citation observations.
guides
Your website may be current while an old report still gives buyers last year's numbers. Audit the downloadable evidence alongside its landing page: check the text, dates, claims, public URL, and revision history before using a citation observation to guide AEO work.
Research section
comparisons
A small team can need enterprise depth; the useful comparison is scope, ownership, and operating burden.
comparisons
Three public-source product profiles, with fit cases and limits rather than an overall winner.
comparisons
Choose a measurement and operating model before comparing AEO platform feature lists.
Research section
benchmarks
A measurement framework for ChatGPT and AI search visibility that keeps implementation, technical eligibility, observed answers, search exposure, and business outcomes in separate evidence lanes.
benchmarks
A defensible monitoring program documents what a run observed, what changed, and what the sample cannot establish.
benchmarks
Seven bounded measures for a documented prompt sample, with formulas, failure modes, and hypothetical examples—not claims of universal engine performance.
Research section
field notes
A four-week operating plan for moving from noisy observations to owned, reviewed changes and a defensible recheck.
field notes
Machine-readable publishing controls can reduce ambiguity, but they do not promise retrieval, citation, or favorable answers.
field notes
A model for thinking about source selection without pretending that opaque commercial systems publish a universal citation algorithm.
Comparison desk
The most recently published public-source comparisons.
comparisons
A small team can need enterprise depth; the useful comparison is scope, ownership, and operating burden.
comparisons
Three public-source product profiles, with fit cases and limits rather than an overall winner.