A firm that sells an audit should be willing to publish the result of running it on itself. That is a low bar. Here is ours.
The number
We ran a combined search, answer-engine and voice audit against harmonygroup.vip. Harmony Group is the same firm as this one, CoCreators Group being its trading name, so the site was ours to score and the result was ours to publish. It scored 12 out of 30.
Eighty minutes later it scored 25 out of 30. Four audits, three fix passes, one afternoon.
The part worth noticing is what did not change. The four practice pages read word for word the same at the end as they did at the start. Nothing was rewritten, nothing was stuffed with keywords and no copy was sacrificed to a crawler. The gap between 12 and 25 was almost entirely machine-readability: what a search engine or an answer engine could extract, and whether it could attribute what it extracted to a named firm rather than to an unattributed page.
Why the two are different jobs
A page can be perfectly clear to a person and close to invisible to a machine. The sentence that says what you do sits inside a paragraph. The firm’s name is in an image. There is no structured description of the organization, no link binding the firm to the people who run it, and no sitemap telling anything where to look.
A human reader fills those gaps without noticing. A generative engine writing an answer about your category cannot, so it uses somebody else’s page instead. Your competitor did not out-argue you. They were easier to quote.
The five points we did not get back
It is 25 and not 30, and the gap has a reason.
The remaining points need a client to give us permission to name them. Specific, named, quantified proof is what moves those last five, and the proof we hold sits inside engagements whose terms decide when a number can be printed. We could have written something vaguer and scored the points. It would have been the kind of claim this whole article is arguing against.
So the score stands at 25 with a stated reason, and it will move when a client says yes rather than when we find a better adjective.
Check it yourself
The site is public and the four practice pages are these:
- Nonprofit strategy and fundraising
- Executive coaching and leadership
- Philanthropy and impact
- Technology and AI
Open one and look at what a machine would find there. That is a more useful test of whether we know what we are talking about than any case study we could write about it, and it does not require you to trust us.
This is also the only proof of this kind that we can currently show you without asking someone’s permission first. When we say a claim is checkable, that is what we mean: you can go and check it, today, without us in the room.