What we will not sell, and why
Every item on this page is a service somebody in this industry actively offers, priced and productized. Some of it the FTC banned in 2024. The rest we refuse on our own. A category where every vendor advertises what they will do for you has nobody writing down what they will not.
The list
Paid or incentivised reviews
Buying reviews, paying for positive sentiment, or offering anything of value in exchange for a favourable one. The FTC's 2024 rule on fake and deceptive reviews bans this outright, along with review gating, which is the practice of screening customers and only routing the happy ones to a review form. If we ever run a review-presence service it will solicit genuine reviews from real customers and accept whatever they say.
Undisclosed community posts
Posting to Reddit, forums or communities on a client's behalf without disclosing the connection. This is sold openly in this industry as a productized service, priced per post, described as building brand profile to influence AI. Under FTC endorsement guidance a material connection has to be disclosed, and beyond the rules it is simply astroturfing. Reddit is roughly 37 to 44 per cent of the social content AI engines cite, which is exactly why it is tempting and exactly why we will not.
Guaranteed citations, mentions or rankings
Engines return different answers to identical prompts. Nobody controls that, including the vendors who sell a guarantee against it, and a promise about a non-deterministic system is a story rather than a guarantee. We guarantee the work, the delivery window and the measurement, because those are the parts that are ours to control.
Placements that hide a material connection
Where we place on third-party pages we favour publishers with editorial standards who disclose sponsored or paid placement properly. A placement that only works because the reader cannot tell it was paid for is worth less to you than it looks, because it is one policy change away from being worth nothing.
Grading our own homework, quietly
We both measure and fulfil, which is a conflict and we are not going to pretend otherwise. What we do about it: the method is published, the prompt basket is fixed and shared so a customer's scan is comparable to anybody else's, the confidence band is reported on every number, and a re-scan that shows no detectable change is reported as no detectable change. If measurement ever stops being able to say our own work did not land, it has stopped being measurement.
What we do guarantee
Narrow on purpose. A guarantee is only worth something if it covers things we control.
- Delivery inside the window stated on the service, which is published in working days rather than described as fast.
- A scan before and a scan after, on the same prompt basket, with the confidence band on both.
- A report that says what the band says, including when what it says is that nothing detectable changed.
A service that could not fail would mean nothing when it succeeded. The measurement has to be able to come back negative, and when it does you will be told, because a vendor whose reports only ever go up is not reporting.
Your results, and who owns them
We publish anonymised measurements from work we do, with your domain and brand removed, unless you opt out. We never publish anything that identifies you without asking first. The full clause is in the terms, and you can opt out at checkout or afterwards, including retroactively.
The services this applies to
Every service carries its own limit, stated on the card, including the ones where the honest answer is that a placement is a rented position rather than a permanent one.