What Do a Coworking Space, a Therapist and a Local Council Actually Have in Common?

Nothing, on paper. But the last four projects I've worked on, a coworking space, a solo therapist, a local council consultation page and a B2B SaaS company, all failed the exact same test before I started, and all improved in the same way once I fixed it. If you're wondering why AI tools recommend a competitor instead of you, this is probably why, and here's the six-point check I now run on every business before touching anything else.

What did these four businesses actually have in common?

None of them had a keyword problem, and none of them really had a content problem either. They had an evidence problem: not enough consistent, credible evidence across the web for an AI system to confidently work out who they were, what they were good at, and why a user should trust them.

Search has changed shape. People no longer type three keywords into Google and pick from ten blue links. They ask longer, more specific questions, across ChatGPT, Gemini, Perplexity and Google's own AI Mode. Google's own VP of Search put it plainly this year, describing the goal of the redesigned search box as helping people ask "the deep, complex or hyper-specific questions that can be hard to articulate," because curiosity "doesn't always fit into keywords." That's not a minor UI update. It's an admission that the old measure of success, does the page rank, isn't the same test any more as: does the system understand the business well enough to recommend it.

What was actually wrong with each of them?

Harbour Collective had loyal members and a strong local reputation, but a Google Business Profile that had barely been touched in two years. Whitfield Therapy had a real, qualified practitioner, but a website that never said so in the way someone asking "who are the best therapists for anxiety near me" actually needs. Elmsworth Borough Council had the right information, buried three clicks deep, so anyone asking a conversational question had to work far too hard to find it. The B2B SaaS company ranked on page one of Google for its key commercial terms and still barely existed in ChatGPT recommendations, because years of blog content had no named author and the company's details varied across more than thirty different directories.

Four different organisations, four different operational problems. Underneath all of them sat the same gap: not enough coherent evidence telling a search system who to trust.

How do I actually check whether AI trusts my business?

I run every business through the same six checks before recommending any changes, because they find the real bottleneck faster than a content audit does.

First, recognition. Ask ChatGPT, Gemini or Perplexity about your business directly, with no extra context. Does it name the right entity, understand your specialisms and location, or confuse you with someone else. If the model can't confidently identify who you are, publishing another fifty blog posts won't fix that.

Second, relevance. A page can be well written and still miss the actual question behind a search. Nobody searches "financial adviser London" any more so much as they ask which independent advisers specialise in people approaching retirement, or which firms work with business owners specifically. Content mapped to the real question wins.

Third, expertise. Advice shouldn't sit on a page as anonymous text. Who wrote it, what qualifies them, and where else are they published or quoted. The person behind the advice matters to a model roughly as much as the page itself.

Fourth, consistency. Your site says one thing, LinkedIn says another, an old directory has your previous address, a press release uses outdated branding. Individually these look trivial. Together they create enough ambiguity that a model can't confidently corroborate who you are.

Fifth, citation. Trust that's built entirely on your own domain is trust a model has to take your word for. A professional body listing you, a publication mentioning you, someone else citing your work, all of that exists independently of your own marketing, and that's exactly why it counts for more.

Sixth, recommendation. Run the actual prompts your buyers would use, "most reputable accounting firms for growing businesses" or whatever the equivalent is in your category, for your brand and your competitors, and see who gets named, who gets described as an authority, and why.

Did fixing this look the same for all four businesses?

The tactics were completely different. The results followed the same pattern.

Business	What I fixed	Result
Harbour Collective	Entity details and local signals	Named in AI answers for "best coworking space" by week 7; tour enquiries up around 38% in 10 weeks
Whitfield Therapy	Named practitioner expertise and schema	Direct inclusion in Google's AI search for key prompts; enquiries roughly doubled over 3 months
Elmsworth Council	Information architecture and retrieval	Residents found the consultation page via AI prompts; response collection ran about three times faster
B2B SaaS company	Directory consistency and named authorship	AI Overview citations grew from 4 of 50 tracked queries to 29; inbound calls tagged "found via AI search" rose from 2 to 11 a month

Doesn't every business need a genuinely different strategy?

Yes, and it's worth being honest about that before this sounds too tidy. A council needed its information architecture fixed, not testimonials. A solo therapist needed named-practitioner content, not a clean-up across thirty directory listings. A SaaS company with a rotating cast of anonymous freelance writers needed authorship reinstated across an entire content library, which a ten-week local project never would have required.

What didn't change was the underlying test. Every AI system, regardless of industry, is quietly asking the same question before it recommends anyone: can I trust this enough to put my name behind the recommendation. The tactics are bespoke. The checklist underneath them isn't.

If you want to see the full detail behind all four of these, the case studies are published in full, with the actual numbers and the clients' own words.

Why do professional services face a higher bar for this?

Because the cost of AI getting it wrong is higher. Ask an AI assistant for a decent local coffee shop and a bad answer costs someone a mediocre flat white. Ask it for an independent wealth manager and a bad answer costs someone real money, or worse advice than they realise they're getting.

In law, finance, accounting and consulting, prospective clients need to verify credentials, regulatory standing and specialisms before they'll trust a recommendation, and AI search has moved that verification step to the very front of the buyer journey rather than removing it. Being crawlable isn't enough in these categories. You have to be understood in the right context, and backed by proof a model can check.

Does this mean my own website barely matters any more?

No, but it does mean your website stops being the whole story. Your site is the one place you fully control, and it still needs to answer the real questions your buyers ask, but AI systems are also weighing everything you don't control: press coverage, professional memberships, named authorship, consistent directory listings, original research other people reference. I think of that second half as the evidence layer sitting underneath Layer 3 of the Search Visibility Stack, AI Retrieval, and it's the hardest part to fake precisely because you don't fully control it.

The practical version of all this is fairly unglamorous. Track the actual questions your buyers ask across ChatGPT, Gemini and Perplexity, and note who gets cited and who doesn't. Favour five genuinely authoritative pieces written by someone real over fifty generic ones. Give your specialists named profiles rather than a generic team page. Clean up your directory listings and professional profiles so they agree with each other. And check your share of the recommendation, not just your rankings, the same way you'd check any other number that actually moves the business.

Why does this matter to me outside the client work?

I'm running the same six checks on something closer to home at the moment, the site for Torpoint Skatepark, a local project rather than a client one. It's a useful reminder that the checklist doesn't care whether you're a global brand or a car park committee that built a skatepark. A parent asking an AI assistant whether there's a decent skatepark nearby deserves as accurate an answer as a buyer asking about enterprise software, and after enough years on a board myself, I've learned the same thing skating taught me: you do the reps until you stop having to consciously think about the check, and you just see the gap.

Where does this leave a business that hasn't checked any of this yet?

Probably where all four of these were before I started, assuming that ranking on Google is the same thing as being recommended by AI, when they're increasingly two different outcomes. The first move isn't a rebuild. It's finding out, honestly, what AI tools currently say about you, if anything, by asking the question your best customer would ask before they ever land on your site.

SEO isn't dying. The scope of what visibility means is expanding, and the goal has quietly moved from ranking the page, to earning the click, to something closer to being recognised. If an AI system can't confidently verify who you are, what you do and why you're trustworthy, it will simply leave you out of the answer.

The businesses winning this aren't the loudest ones. They're the easiest to understand, the easiest to verify, and the easiest for a model to confidently recommend.

Search visibility was never just about being found. Increasingly, it's about being recognised.