Case studies, published in full.

What you will not find on this page.

No client names. No logos. No anonymised client whose industry, city and headcount narrow it to one business. That is a rule rather than a gap in the paperwork, and it does not change when a sale depends on it.

So the evidence here is a different shape. It is our own work, run in public, on categories nobody commissioned, published whole and left alone afterwards. You can check every part of it without asking us for permission, which is not true of a testimonial.

September 2026. Electricians, West Palm Beach, FL.

Run 24 September 2026, published 14 October 2026.

13 businesses were named across five answers.

2 were named in all five: Mister Sparky of West Palm Beach and Grizzly Electric. Those two had the category to themselves.

5 were named once only.

An independent check was run the same day, and the entry carries a figure we withdrew because we could not reproduce it from what we had kept.

What it showed

The withdrawn figure is the part worth reading. It is published inside the entry, under its own heading, and the entry has not been edited since. A result nobody can argue with is usually a result nobody checked.

Read the run in full →

The first ninety days.

Written out so you know what the quarter looks like before you pay for it, and so you can hold it beside anybody else's plan. We run this for businesses across the United States. The two runs published above are local categories because that is where the five questions are sharpest, not because the work is local.

Days 1 to 14. The baseline run.

We put the five buying questions to the assistants for your service and your area, and write down every company each answer named. In the same fortnight we read your website the way an assistant reads it: whether the crawlers are allowed in at all, what your structured data says your business is, and whether your name, address and profiles agree with each other across the places that carry them.

Days 15 to 45. Fix what the engines read.

Crawler permission first, because writing for assistants that cannot open the page is the most expensive mistake available. Then entity consistency across every profile that carries your name, the structured description of your business, and the pages that answer your buying questions in words rather than in pictures.

Days 46 to 75. Build the evidence they cite.

Answers reward third-party evidence: reviews on the places buyers already trust, profiles that agree with each other, and published material an engine can quote. This is the slowest stage and the one that needs you, because reviews come from your customers and nobody else can ask them for you.

Days 76 to 90. Run it again and compare.

The same five questions, the same shapes, the same assistants. You get the second run in full, set beside the first, including every place where nothing moved.

What we do not promise is a seat in an answer by day ninety. Answers move by question, by platform, by location and by day, and anybody promising you a place in one is selling something they do not control. What we do promise is the same measurement before and after, and that what failed to move is published next to what did.

Questions buyers ask about this.

Do you publish client case studies?

No. We do not name clients, publish their figures, or describe one closely enough that you could work out who it is. A business that hires us to be named in AI answers is handing us its reputation, and the day that becomes marketing copy it is worth less than the copy. What we publish instead is our own work in public, on categories nobody paid us to run. If you want to know whether the work is any good, read a run and then ask us to put the same five questions to your category.

Who paid for these runs?

Nobody. No company commissioned either run and no business named in one was told it was happening. Both entries say so in their own first paragraph. It matters because a run somebody paid for is a run somebody could have shaped, and the only value a published result has is that it was not arranged in advance.

What is actually inside a published run?

The five questions exactly as they were asked. Every company named under each question, in the order the answer named it. How often each name appeared across the five. The sources the answer leaned on. The date. And a section listing what we did not record, which is the section most reports leave out. Both entries also carry the limits in plain words: one run, on one day, through one engine.

How do I know the results were not picked to flatter you?

Ultimate Elite Solutions is not named in either run, because neither run asked about our category. The questions are fixed shapes with only the service and the place changed, so one run can be compared against another, and they are published alongside the results. Entries are never edited after publication. The September run carries a figure we took back, under its own heading, because we could not reproduce it from what we had kept.

Can I see the plan and the numbers before I hire you?

Both are on this page, above. The first ninety days are written out in four stages, and the list of what gets reported is written down beside them. If an agency cannot show you both before you pay, you are being asked to buy a result rather than a method. Ask us for the same thing in writing and you will get this page back, unchanged.

Can you run this for my business?

Yes, once, for nothing. The free report puts five buying questions to an AI answer engine for your service and your area, and sends back the answers themselves: every company each one named, in order, with the date and the sources. If your business is there you can read the sentence around your name and check whether it is right. If it is not there, you can see who is.