What we don't do
This is a young market, so plenty gets tried. Some of it works for a while, some never works, and some costs you your credibility further down the line. Below is what we don't do, with the reason behind each one. Feel free to put the same question to whichever agency you're talking to now.
- hidden instructions for language models
- bought or invented reviews
- fake accounts on forums and communities
- guarantees on positions in AI answers
- promises on measures without evidence
Five things you won't get from us. The reason is below, point by point.
Five things you won't get from us
No hidden instructions for language models
People experiment with text meant only for a model to read: white letters on a white background, hidden lines in the metadata, blocks made invisible with CSS, or a sentence that tries to tell the model which company to recommend. We don't take part in any of that. It misleads the reader, it falls under the spam policies of search engines, and it gets recognised more easily every year. The damage lands the moment your site gets kicked out, and then the gain of a few months is gone.
No bought or invented reviews
Reviews carry a lot of weight in what AI says about you, so the temptation to buy them is real. We don't write reviews for you and we don't buy them in. What we do is ask you to reach out to your satisfied customers, set that process up properly, and make sure your profiles are complete and easy to find. It's slower, and it holds up.
No simulated presence on forums and communities
Reddit and specialist forums are important sources for AI answers. Some agencies use invented accounts there to recommend their own client. We don't post under a false name and we don't stage fake conversations. If you want to be mentioned in communities like that, we do it through real contributions, or through the people who already belong there.
No guarantees on positions in AI answers
Nobody can promise that ChatGPT puts you in first place. Where you land in an AI answer varies by run and by day. The odds of two identical questions producing the same list of brands are small. Anyone who guarantees you a position is selling you an outcome he doesn't control. So instead we commit to something we can actually control: whether you're mentioned on twenty agreed prompts, measured over thirty days. That we can steer, and that we can check.
No visibility promise on measures that don't support one
This is the hardest of the five, because it's about our own trade. Two examples. Schema markup gets sold everywhere as the key to AI visibility, while the only controlled study that isolated its effect found no measurable rise in citations. And llms.txt gets presented as the file the models read, while a measurement across 137,000 sites found that 97% of those files were never fetched. We set both of them up properly, because they cost little and can't do any harm. We just don't attach a visibility promise to them, and we don't put them at the front of the plan. What goes at the front is what's actually proven: access for the crawlers, server-side content, speed, indexation and mentions on independent sources.
You're buying something you can barely check yourself
With AI visibility, you're buying work you can only half check. You see the invoices and you see a report. Whether the method itself is sound, you largely have to take on trust. That's why I'd rather put my own limits on paper beforehand, so you can hold me to them.
It's useful to you too when you're talking to other agencies. Ask them what they don't do. Ask whether they guarantee a position in an AI answer. Ask where the evidence is for the measure they want to carry out first. You'll hear pretty quickly, from the answers, who takes the trade seriously.
See how we do measure it, in the Findori Visibility Framework →
Want to see where you stand first?
Do the free check or book an audit. Then we talk about numbers, not promises.