Anything Engine Optimization The rolling record of the AI-search industry.

Accountability

What this record is made of

Every publication asks readers to trust its judgment. Almost none shows its working. These figures are counted from the published record when the site builds — not entered by hand, not rounded in our favour, and not excluding the ones we would rather were different.

520 items over 119 days · 4.4 a day · 131 distinct sources

01 Where it comes from

Concentration is the number to watch. Our founding memo claims the wire catches stories before or without the trade press; the honest reading of this table is that it does so less often than that claim implies.

SourceItemsShare
Search Engine Roundtable7214%
PPC Land6212%
Search Engine Journal438%
Chat GPT Is Eating the World296%
Profound research blog234%
Search Engine Land224%
arXiv GEO214%
BrightEdge Blog163%
The four largest trade outlets together supply 199 of 520 items (38%). A wire that mostly follows the trade press is an aggregator; moving this number down is an editorial goal, and publishing it is how it stays one.

02 What we flag against ourselves

A flag is a caveat we attach to our own item — that it rests on one source, that the study was published by a vendor with an interest, that the news is older than it looks. 272 of 520 items carry at least one (52%).

CaveatItems
single-source78
catch-up29
BrightEdge study17
Profound study10
Semrush study5
Ahrefs study5
spotted via documentation change5
Sistrix study4
arXiv preprint4
vendor-published3
SE Ranking study3
vendor study3
Seer Interactive study3
vendor announcement3
single-probe3
Similarweb study2
Peec AI study2
Fractl study2
Adthena study2
self-published analyst data2
wire analysis of published documentation2
Comscore data2
Scrunch study2
research study2
wire's own measurement2
audit study1
documentation change bounded to Jul 23–24 UTC1
Google study1
Pillarbase study1
Writesonic study1
geoSurge study1
AOP-commissioned study1
Datos/SparkToro study1
Search Signals Index1
content recency study1
RTB House study1
local AI search study1
Norly Research study1
single-brand self-study1
Promptwatch study1
rolling out to a subset of site owners1
wire analysis of the protocols' own published documentation, fetched 2026-08-301
wire analysis of Cloudflare's published posts1
wire analysis connecting two published measurements1
OpenAI-reported figure1
SALT.agency audit1
NewzDash analysis1
SISTRIX study1
Contentsquare study1
Sensor Tower study1
Productrise study1
Ten Speed study1
Smarter Ecommerce study1
AlsoAsked study1
Similarweb data1
AI Overviews CTR study1
Ahrefs data1
Ahrefs analysis1
Comscore study1
SparkToro/Datos study1
citation-drivers study1
preprint study1
AI Overviews branded-search study1
LintLab study1
Recrawled study1
invalid-click study1
DemandSphere study1
HasData study1
ProxyPicker study1
Typeface study1
Chartbeat 2026 Publisher Playbook1
CommerceV3 study1
study1
BrightEdge data1
OpenAEO study1
DataDome study1
PRecious Communications study1
Cloudways study1
TopRank/Ascend2 study1
Determ study1
DataReportal report1
Searchable study1
Truelogic study1
ContextFetch study1
AI Overviews brand-search study1
Analyst study on vendor data1
Vercatus study1
Graphite study1
Adthena data1
TJ Digital study1
spotted via Search Engine Roundtable1
advocacy1
McKinsey study1
OpenAI and Anthropic positions via secondary reports1
Wikimedia Foundation is an interested party1
Kaiser & Schulze working paper1
spotted by SEOs1
SearchDock study1
Google detail via partner statements and a Google VP's post on X1
Profound product announcement1
Adthena report1
OpenAI's own figure1
StackScan study1
single-site case study1
test1
SparkToro/Similarweb study1
Peec AI survey1
Flags are set at publication and never removed quietly.

03 Corrections

27 of 520 items carry an appended, dated correction. The standing log is at corrections.

04 Coverage and furniture

MeasureCountShare
Items answering a searchable question520100%
Items tagged to the record45988%
Subjects with a rolling record page25—
Lanes in use9 of 9—
Items with no entity tag are absent from the rolling record — a gap in our own filing, counted here rather than left for someone else to find.