Research · what the notes leave open

Open questions

Each field note is a single observation. Each one raises a question it does not answer. They are listed here so it is clear what is known and what is not, and which study would have to come next.

• Q01How many repeated checks of a fixed question are needed before movement against a baseline can be told apart from session-to-session drift?from Ask the same question twice, get two answers
• Q02For a given category, which of the content-quality changes the paper tested shift AI answers, and by how much?from What the research says actually moves AI visibility
• Q03Does the Reddit and arXiv dominance hold for each engine separately and on a repeat run, or is it carried by one engine's citation volume?from Ask AI about AEO, and it cites Reddit
• Q04Does each engine's willingness to accept a false premise stay the same across sessions and repeated runs of the same leading question?from Five engines refused to make something up. One didn't.
• Q05After a company earns a place in its category's most-cited domains, how long before assistants' answers in that category cite or name it?from Where AI actually looks for your category, and how to be there
• Q06How much consistent third-party description does a new name need before engines resolve it to the intended company rather than a namesake?from Six engines, five Makefields
• Q07Within one named category and a fixed question set, which common checklist items produce a measurable change in AI answers?from Why most AEO checklists are wrong
• Q08After a source page is corrected, how many weeks pass before each engine's answer reflects the change, and does the lag differ by engine?from AI got a fact about you wrong. The order of operations to fix it.
• Q09In one category, what share of links cited by AI answers resolve, and what share of those support the claim they are attached to?from AI cites sources that don't check out
• Q10How much of the 83% single-engine share comes from each engine's long tail of follow-up sources rather than its main citations?from Do AI engines cite the same sources? Barely.
• Q11When Google AI Mode shows no visible sources, are sources absent, or presented in a form the capture method does not record?from Some AI engines cite 14 sources per answer. Others, none.
• Q12Does the higher agreement on more specific questions hold when the same seven questions are re-run on several different days?from Six engines, six different 'best' agencies
• Q13For engines that only search when asked in the text, how much does the added instruction change their answers compared with the unaltered question?from Why I never type 'search the web' into an AI visibility test
• Q14For an engine that requires sign-in, do answers with memory and custom instructions off match the answers it gives signed out?from The ghost in the logged-in answer
• Q15In one category, how often does a page that ranks in Google's top ten appear among the sources an AI answer cites?from Cited, not ranked: the shift SEO people keep missing
• Q16For one company over time, do the four readings move together, or can one improve while another gets worse?from How to read your own AI-visibility audit
• Q17In one category, is Reddit's share of citations higher for opinion-shaped questions than for factual ones?from Why AI leans on Reddit, and what to do about it, honestly
• Q18Over repeated runs, does a rise in a company's citations come before a rise in its recommendations, or do the two stay unrelated?from Recommended ≠ cited: two different games
• Q19Which of the three steps (consistent account, machine-readable identity, outside corroboration) changes engines' recognition of a new name first?from The entity-recognition mountain every new brand starts on
• Q20Across a fixed question set, how often do engines answer a different reading of the same question, and does this vary by engine?from AI engines don't even agree on what the question is