The 16% Synthetic-Source Finding Depends on an AI Detector
A 712-query audit found detector-classified synthetic pages among citations, but 27.1% of cited URLs were not analyzed and page accuracy was not tested.
From this desk
Start with the most useful page
The featured page offers a strong introduction. The remaining cards help readers move through this editorial section without turning the archive into a service menu.
Read the flagship case study
How SearchEngineAnswer went from zero to early search visibility
This first-person build record connects publishing decisions with indexing evidence, Search Console observations, limitations, and the next measurement window.
Case study
Connect decisions with observed outcomes
Follow the site build, indexing evidence, changes, and limits instead of treating one visibility event as a universal rule.
Experiments
Change one material variable
Preserve the baseline, observation window, confounders, and decision rule so a result remains inspectable.
Evidence
Keep every claim close to support
Classify facts, observations, interpretations, recommendations, and opinions before drafting conclusions.
Latest research
Completed tests, case studies, and evidence reviews
Each research article identifies the question, evidence, method, result, limitation, and boundary on what can be inferred.
-
-
Most AI Citation Sources Appeared Once in a 16,069-Citation Sample
A source-long-tail recomputation plus a separate LinkedIn engine-mix example show why aggregate AI citation rankings can hide most domains and platform dependence.
-
Google AI Mode Cut Publisher Click-Through by 18.8 Points in a Field Experiment
A field experiment found AI Mode assignment reduced publisher click-through by 18.8 points. Separate ITT, LATE, limits and site-specific measurement.
-
What 3,602 Early ChatGPT Ads Reveal About Delivery
A 91-account audit recorded 3,602 ChatGPT ads. See the delivery rate, count discrepancies, income finding, limits and audit ledger.
-
What 7,473 AI Citations Can and Cannot Tell Us
BrandGhost’s latest Observatory run offers useful signals about citation age and engine overlap, but its date and source-category coverage materially limit broader claims.
-
Are People Also Ask Answers Becoming AI-Generated?
A third-party report says AI-generated People Also Ask answers reached 97% of its English-language sample. Here is what the number can support and how to audit it.
-
Agentic Commerce Standards Split Across Protocols, Data and Trust
The W3C commerce workshop exposed a stack, not one standard: transaction protocols, product data, identity, consent, payment and trust still need to interoperate.
-
Perplexity Citation Audit: When a Link Does Not Support the Number
We recalculated two public Perplexity citation files, separated row and claim denominators, and built a four-state audit for numerical source support.
-
Two GEO defense papers test different failures in AI answers
Two September 2026 papers study GEO defenses. Compare factual distortion with source-selection manipulation, and download our evidence-review worksheet.
A useful standard
How to read SearchEngineAnswer research
A useful study makes its conditions visible. Treat each result as evidence within a declared sample and environment, not as a universal platform rule.
Method before conclusion
The question, inputs, environment, measures, and stopping rule should be understandable before the result is interpreted.
Failures stay in the record
Timeouts, absent results, rejected cases, and contradictory observations are data, not material to hide.
Limits remain visible
A measured association or one platform observation does not automatically establish causation or a permanent ranking rule.
References and next steps