Perplexity Photon Speeds Retrieval but Does Not Explain Citations
Perplexity’s Photon infrastructure and Fast Search mode change retrieval speed and cost. Neither explains which sources become citations, so a paired benchmark is still required.
Tagged collection
Latest articles tagged “AI Citations”
Open an article for the complete method, evidence, limitations, and a focused reader discussion.
-
-
CITECHOICE Study: More Citation Credit, Uncertain First-Citation Gains
A controlled AI-search replay found more citation markers for structured source text, but no clear gain in whether a source was cited at all. Here is what publishers should measure.
-
AI Answer Citations Can Fail at Four Separate Layers
Four recent papers expose separate failure points between retrieval and a trustworthy AI citation. SearchEngineAnswer reproduced PageRecall’s public tier-0 checks and built a claim-level audit ledger.
-
AI Search Engines Cited Different Websites in a 960-Answer Study.
A 960-answer experiment and a separate B2B panel show why AI citations vary by engine, prompt type and measurement unit. Their percentages should not be pooled.
-
CiteShade Raised Wrong Answers From 1% to 68%
A multi-source RAG attack made incorrect answers cite trusted evidence. Source-deletion tests reveal which document actually drove the claim.
-
CITECHOICE Finds Structure Shifted Citation Credit, Not Source Admission
A causal replay found that structured rendering increased citation count, while its effect on whether a source was cited at all remained inconclusive.
-
ChatGPT, Claude, Grok and DeepSeek Search the Web Differently
A 171,264-conversation study shows that four AI platforms differ at every search stage, from deciding to browse through selecting citations.
-
What 30,000 Reddit Answers Reveal About Which Voices Survive AI Search
A 30,000-output study finds retrieval drives most variation while Reddit Answers favors early, top-level and more formal comments in its selected evidence.
-
Q2D-Web Shows Why AI Citations Depend on First-Stage Retrieval
A 190-million-document benchmark shows why retrieval recall, unique evidence coverage and downstream citations must be measured separately.
