Three Businesses, Three Trust-Signal Scores, One Fix Order

Reading time: 10 minutes

Last verified: 11 August 2026

One business went from 7 out of 19 to 14 in twelve weeks. One went from 4 to 9 and is still climbing. One went from 8 to 12 to 16 across a year. Same scorecard, same first three fixes, three very different sites.

This is the article I wanted twelve months ago, when I was still guessing which trust signals mattered most. Rather than theory, here are three real businesses we have worked with: what each scored on the 19-signal trust scorecard, the order we fixed things in, and how far the score moved.

I need to be straight about what this post is. All three have asked not to be named, so they are not named here, and every revenue figure has come out with them. What is publishable is the pattern, not the profit and loss. That is the better half of the trade anyway: a revenue number from a business you cannot identify is a number you cannot check, while a score and a fix list are things you can run on your own site this week.

→ The 19 signals are not academic. They are the difference between being cited by an AI answer and being invisible to a buyer with their wallet already open.

This covers all three side by side: the same framework across three categories (plumbing, retail, interior design) and two countries. By the end you will know how the scorecard works, which three fixes we always do first, why a bigger site takes longer to work through, and what a realistic first phase looks like.

How do you score a service business against the 19 trust signals?

Quick Answer: Five categories of signal: who you are, what you publish, how you are reviewed, how the rest of the web talks about you, and your structured data. Each signal scores one point if it is present and verifiable on the live site, and zero if it is not.

The scoring is binary on purpose. The audit surfaces what you can prove today, not what you used to have or intend to have. No half marks, no grading on a curve.

The 19 signals are the Ambitions scoring framework. We built it from the editorial patterns we kept seeing in the sources that AI answers do cite: named humans, dated pages, clean structured data, a live review feed, third-party mentions. Nobody outside the AI labs knows what a model checks internally, and I will not pretend otherwise. What we can do is score the things a model can see, the same way at the start and again at six and twelve months. For the three below, the method was identical. Only the pace changed.

How did the US plumbing company go from 7 to 14 in twelve weeks?

Quick Answer: An established plumbing company in a US metro area, with twelve service pages. It started at 7 out of 19 and now scores 14. The whole move happened across twelve weeks of editorial fixes, in a strict order.

A note to avoid confusion, because there is a second plumbing business in this series: this is not the $5,000 MRR business I write about elsewhere. That is a different company at a different stage, with a different score and a different problem. This one was established and trading well when we audited it, and the trust-signal layer was almost entirely missing. Reasonable website, good Google reviews, and nothing telling an AI engine who to credit for any of it.

What we fixed, in order:

Weeks 1 to 3: a named author on every service page. The owner and the lead plumber, both with photos, short bios and their qualifications.
Weeks 4 to 6: Service schema rolled out across all twelve service pages, plus FAQPage schema on the top three.
Weeks 7 to 9: review automation switched on, so new reviews arrive on a standing cadence rather than in occasional bursts.
Weeks 10 to 12: three citation pieces. One trade press article, one podcast appearance, one local roundup.

The score moved from 7 to 14 across those twelve weeks.

Twelve weeks is fast, and the size of the site is a large part of the reason. Twelve service pages is a small, tidy architecture. Every fix could be applied by hand and checked in Google's free Rich Results Test in a single sitting, with nothing left half done while the rest caught up. Hold that thought.

Why did the US luxury bed retailer only reach 9 out of 19?

Quick Answer: It started at 4 out of 19, the lowest of the three, and now sits at 9 and still climbing. The fixes were the same. The site has over 200 product and category pages, so the same fixes were a much bigger job.

I covered this one in detail in "Four months of flat dashboards, then the work finally showed up". Starting score 4. After phase one, which was schema, named authors on category pages and review automation, it was at 9. A five point move from the worst baseline of the three.

Site size is the whole story of the difference in pace, and the reason is simple arithmetic rather than anything mysterious. On twelve service pages, rolling out Service schema is an afternoon. On 200-plus product and category pages, it is a templating job and then a QA job: a template per page type, edge cases caught by hand, and a long tail of pages that each have to be reached before the work is actually on them. The score measures what is verifiable on the site right now, so a change that is only half rolled out only half counts.

Reviews were running at four a quarter before we started, close to invisible for a retailer of that size. That was the signal with the least standing in its way.

I include this one because not every score lands at 14 in the first phase. A 9 is a real result, on a harder site, from a lower start.

What does a 16 out of 19 look like, and how long did it take?

Quick Answer: A UK residential interior designer, working with buyers aged 45 to 65. She came in at 8 out of 19, reached 12 by the end of quarter one, and finished year one at 16. Twelve months of steady work, not a sprint.

This is the case I show owners who think their category is too crowded to bother. UK residential interior design is busy, but when we audited her competitors the trust-signal layer was thin across the board. Across twelve months:

Named as the author on every service page, with her credentials stated plainly, plus a senior designer credited on project case studies.
Service and Article schema rolled out across the page architecture.
A Featured In section built from three UK trade press placements.
Review automation switched on, so the feed stays current instead of ageing.
Four short case-study videos filmed and published, with clients on camera.

The score moved 8, then 12 at the end of quarter one, then 16 at the end of year one. Her buyers are 45 to 65, an audience plenty of people still assume does not use AI search.

The jump from 12 to 16 took three quarters, and none of it came from the site itself. It came from press placements, video and the slow accumulation of third-party mentions. The cheap points go first. The last few are earned.

What pattern do all three share?

Quick Answer: All three were missing the same three foundational signals at the start: a named author, service schema and a live review feed. All three moved most from fixing those three first, then compounded it with press, podcasts and citation pieces in months four to twelve.

The fix order was identical across three categories and two countries. First:

A named author on every service page.
Service schema across the page architecture.
Review automation, so the feed is current rather than historic.

Then, in months four to twelve:

Citation pieces in trade press, podcasts and local roundups.
Founder-led content on the channels where the buyer actually searches.
A year-end rescore with a forward fix list.

Three businesses is a small sample and I will not dress it up as research. But the same three gaps were open in all three, and closing them in that order produced the largest part of the score move every time. That is a strong enough hint to act on when the fixes are within reach of a determined weekend.

Liyana's insight

I would rather publish a score than a revenue figure. Revenue is the thing everyone wants and the thing nobody can check, and I have watched our industry lean on numbers that dissolve the second you ask a question about them. A score is built from things you can look at on a live website: is there a named human on the page, does the structured data validate, is the newest review from this month or from 2023. You can score a competitor without their permission and score yourself without ours. That is why taking the money out of this article cost it much less than I expected.

How do I find out what my own score is?

Quick Answer: Run the 5-Step Audit on yourself. It is free, it takes about half an hour, and it gives you the same binary read we use: present and verifiable, or not.

If your score comes back in single digits, you already know your fix list: named authors, then schema, then a live review feed. If you can do that over a long weekend, do it and keep your money. We exist for businesses that want it done faster and paired with the citation and press work that takes months to build at scale. A Diagnosis runs £497 to £1,500, and a Framework Build £5,000 to £8,000.

Comment AUDIT on any Ambitions AI post on Instagram or LinkedIn and ManyChat sends the 5-Step Audit Checklist to your DMs.

So where does that leave your business right now?

All three started between 4 and 8, and each picked a different pace. The plumbing company moved fast, over twelve weeks. The retailer staggered the work across six months, because its site made that necessary. The designer went all in over a year.

The fourth path is to wait, watch, and assume AI search is hype. It is the only one of the four where nothing on your scorecard changes, because none of these signals appear on their own.

Pick one signal this week. Put your name and face on your main service page, with your qualifications written out. That is one point, and it is the point everything else hangs from.

FAQ

Without names or revenue figures, what is this proof of?

All three asked not to be named, and that is their call. What is left is evidence that a specific, repeatable fix list moves a measurable score across different categories and site sizes. It is not evidence of what that score is worth to you in pounds, and a named figure would not have proved that either. Score movement is something you can verify on your own site. Take the fix list, not my word.

Why did the retailer's score move slower than the plumbing company's?

Size. Over 200 product and category pages against twelve service pages. Schema and author work has to be templated and checked across the whole architecture, and until it is actually live on a page, that page scores nothing. Same pattern, longer timeline.

Will the same fix order work for my category?

For most service businesses, yes. A named author, service schema and a live review feed are sensible first steps almost anywhere, because they are what a machine can read and verify. The press and citation work in months four to twelve looks very different by category.

What is the minimum score I need before AI engines cite me?

There is no published threshold, and I would distrust anyone who gives you one. Nobody outside the AI labs knows what a model weighs, so any number offered as a citation cut-off has been made up. The useful question is not what number gets you in, it is whether a page gives an engine anything to credit: a named human, structured data that validates, a review feed that is current. A page missing all three gives it nothing to work with, whatever the score on it.

Is the free audit the same thing as a paid Diagnosis?

No. The 5-Step Audit is a self-serve checklist you run on your own site in about half an hour. A Diagnosis is work we do, at £497 to £1,500, and it goes wider than the checklist. Start with the free one. It costs you nothing and it may be all you need.