THE ANSWER REPORT
01. OVERVIEW
The Answer Report is my own product. It measures how well an AI assistant can read a business website, publishes the score, and hands over the fix.
People increasingly ask an assistant who to hire before they ever open a search box, and most websites give those systems almost nothing to work with. There was no honest instrument for that, only opinions and vendors selling a feeling.
So I built one. A deterministic engine that scores a site out of 100 across the things an assistant actually reads, an editorial layer that publishes those scores as a newspaper, and the rails that keep the whole thing running with nobody standing next to it.
02. THE CHALLENGE
A product that grades other people has to be unarguable, so almost every design decision here is really a decision about trust.
03. THE SOLUTION
One product in three layers: the engine that measures, the publication that carries it, and the automation that runs both.
The engine, and the rule that keeps it stable
A weighted set of checks covering crawler access, structured data, plain-text legibility, contact details and more, resolved into a readiness score out of 100. Run the same site twice and the number is identical, because the checks are fixed and the pages sampled are chosen by rule rather than by chance.
A newspaper, not a dashboard
The findings publish as an actual publication: volume and issue, a front page, an edition per cohort, findings pieces, and a methodology page in plain language. A masthead makes a claim a dashboard cannot. This is on the record, and it is signed.
The machine view
Any page on the site flips, by a control or a single keystroke, into exactly what an AI reader receives: the served structured data, the crawler directives, the text skeleton. The product demonstrates its own thesis on its own pages.
An open audit desk that cannot be abused
The free audit runs the same deterministic engine, live, against the address you file. Only public addresses are accepted and private or internal ones are refused outright, requests are capped per visitor per day, and one job runs at a time behind a hard timer. There is no path from a visitor to the expensive part of the pipeline.
The proof, made portable
Every report gets a score badge served as an image, addressed by a one-way hash of the report token, so the token itself never leaves my side. A business that fixed its score can hang the number on its own site.
It runs itself
Auditing, publishing and delivery run as scheduled jobs on their own server, with a watchdog that raises an alarm the moment serving looks wrong. The standing rule behind it: nothing that earns money is allowed to depend on a machine at somebody's desk.
04. WHAT IT DOES NOW
The engine grades itself in public, and the findings have started doing the arguing:
It publishes its own score
The engine gives the publication 93 out of 100. The missing points are a phone number it does not have, and rather than invent one to round the number up, the correction is printed on the front page.
The findings are part of the product
Of the 173 businesses probed, 153 were named by neither assistant, while a pair of national directories out-named every real business in the set. That is published, with the method, for anyone to check.
One insight worth the whole build
A default setting on a very common hosting layer quietly blocks AI crawlers and hides the contact address behind obfuscation. Turning it off took our own studio site from 60 to 93. Most owners have no idea it is on.
It grows by edition, not by release
Adding an edition is a data step rather than a development one, so the publication expands without a deploy standing behind it.
I wanted an instrument, not an opinion. If a score is going to cost someone money to fix, it has to come out the same every time and it has to show its evidence, starting with its own.