Corrections
Every change to a score or a claim on this site lands here, dated,
with what was wrong and what replaced it. Scores are reviewed quarterly. The next review is
12 December 2026.
A directory that ranks agencies is going to get things wrong, and the ones
that do not keep a log are not the ones making fewer mistakes. Three entries are already here on
the day this page went up.
15 September 2026
Score corrected
Score moved
TradeOps Consulting republished its measurement method, cadence and limit
An earlier read scored TradeOps Consulting zero on measurement method,
citation change over time and falsifiability because those pages did not state a prompt set,
a run cadence or a failure condition. The AEO service page now publishes all three, so the
scores were re-read against the same five criteria.
Method moved from 0 to 5, change from 0 to 3, entity from 4 to 5, falsifiability from
0 to 4, and engines from 4 to 5. The weighted total is now about 93, which places TradeOps
first on measured score rather than eleventh. The editorial placement that used to pin it
first while scoring low has been removed; the row uses the normal Score stamp.
What changed on their side: a fixed HVAC prompt set sized by service and city, monthly
re-runs with period-over-period citation share, and a stated three-cycle failure condition
when share does not move. The previous standing entry that recorded the three zeros is
superseded by this correction.
12 September 2026
Score corrected
Our error
Plumber SEO was filed as an explainer and is a full service
Our first pass placed Plumber SEO in the tier for companies that publish
answer-engine articles without selling answer-engine work. Reading the page properly shows a
full service offering, and the strongest falsifiability position of any company we have
scored.
The page carries a section headed with what the company will not tell you: that SEO is
dead, that it can guarantee ChatGPT will name you, or that the shift is urgent enough to
skip the fundamentals. On the guarantee it says plainly that these systems do not publish a
ranking algorithm and their answers vary by phrasing, location and session. It also declines
to bill answer-engine work as a separate line item, on the grounds that the signals overlap
almost entirely with work the client is already paying for.
Falsifiability moved from unscored to 5 of 5. The company moved from the bottom tier to
joint fifth. The mistake was ours and it was caught on a second read rather than reported
to us.
12 September 2026
Count corrected
Our error
We said sixteen companies. It is at least eighteen
Our first count of companies selling answer-engine work to home services
put the figure at sixteen. A second search pass found two more with full service pages, one of
which turned out to publish a genuine measurement method. The honest figure is at least
eighteen, and the true number is unknown.
Unknown rather than eighteen, because two companies block our reading of their pages
entirely and we will not score a company from a search result. One of them has a snippet
suggesting it names six engines, which would place it high on this board. It stays unscored.
The lesson we are keeping: a set that looks complete after one pass usually is not, and
the second pass is where the company that changes the conclusion turns up.
How to get something corrected
If a score on this site is wrong, the fastest route is to point at the page
that proves it. Scores come from published material, so a page that already states a prompt
set, a cadence or a failure condition will move a score as soon as we have read it. Every
change lands here with the date and what was wrong.
We will not remove a company for asking, and we will not raise a score in
exchange for anything. What we will do is re-read the page and publish the result either way.