E-E-A-T in Action: A One-Man Locksmith's Two-Month Google and AI Results

Back in April we wrote about a Brewood locksmith, Lockerfella, that hit number one on ChatGPT and Gemini for its home-town search within weeks of launch. The fair question came straight back: nice for a niche term, but can a brand-new one-man site actually compete for the searches that bring in money? Two months on, we have the data. Lockerfella went live on a brand-new domain on 8 April 2026. By 21 June it ranks on page one of Google for "locksmith wolverhampton", tops the map pack in its home towns, leads Google's AI Overview, and gets recommended unprompted by ChatGPT, in one of the most competitive and spam-ridden corners of UK local search. In one revealing case, it doesn't. This is what two months looks like across four different surfaces, and the exact build decisions that earned it.

What two months on a brand-new domain actually looks like

Two numbers frame everything that follows. The site is two months old, on a domain with no history, no aged authority and barely a backlink to its name. And the locksmith trade is brutal. Emergency locksmith searches are high intent and high value, which is exactly why the niche is swamped with national lead-generation networks, call-centre operations and a long, grubby history of fake-address listings. This is not a quiet village category where a tidy website wins by default. Ranking a real one-man firm above that noise is hard, and that is the whole point.

Here is where Lockerfella sits as of 21 June 2026, checked from a West Midlands location in a clean browser. Search positions move, especially the local map pack, so treat these as a dated snapshot rather than fixed facts.

Search	Google organic	Google map pack	ChatGPT (search on)
Locksmith Penkridge	#6	#1, plus knowledge panel	Top recommendation
Locksmith Codsall	#2	#2	Recommended
Locksmith Brewood (home)	Strong	Local	Top recommendation
Locksmith Wolverhampton	#5 (page one)	Not in pack	Not named
Locksmith Stafford	#5	Not in pack	Mixed
Locksmith Cannock	#8	Not in pack	Not named
Locksmith Birmingham	Out of core area	Not in pack	Correctly not chosen

Read that table twice, because the interesting story is in the disagreements between the columns, not the wins. A page-one Google ranking and a ChatGPT recommendation are not the same thing, and they do not always land together. We have spent the last two months watching them diverge, and the reasons turn out to be a clean lesson in what E-E-A-T and non-commodity content actually buy you.

Does ChatGPT recommend local businesses? How we tested it

ChatGPT does recommend specific local businesses. Ask it for a locksmith in your town and it will narrow the field to two or three named firms rather than hand you a directory. So the question worth answering is not whether it recommends, but who it picks, and why. We tested that as cleanly as we could.

Every query went into a fresh Temporary Chat, which switches off memory and personalisation, so nothing in our own history could nudge the answer. We led with blind, customer-style prompts that never mentioned the brand: "Who's a good locksmith in Penkridge?", "I need an emergency locksmith in Wolverhampton tonight", "Who does uPVC door lock repairs around Codsall?". If Lockerfella surfaced, it surfaced on its own merits. One honest caveat, stated up front because it changes how you read the results: web search was on for every test, shown by the source citations under each answer. That means ChatGPT was reading the live web, not just its training data. So in part we are measuring the same public signals Google sees, filtered through ChatGPT's own judgement about who to trust. We also ran deliberate controls for towns outside his core patch, to check the model was reasoning rather than just flattering our client.

Where he wins on both Google and ChatGPT

Closest to home, the two systems agree, and they agree strongly. In Penkridge and Codsall, the towns nearest his Brewood base, Lockerfella is in the Google map pack at number one and number two, holds organic listings, and is the business ChatGPT reaches for first.

That last line in the screenshot is the whole thesis in miniature. ChatGPT did not just name him; it gave a reason, and the reason was the quality of the page. It is rewarding content that reads as written by someone who actually does the work. Ask it the broader question and the pattern holds.

It is not only the generic "best locksmith" query, either. When the question gets specific, about a particular job in a particular pair of villages, he still comes up as the first name.

And the underlying Google picture matches. In Codsall, he sits at number two in both the map pack and the organic results, a dual presence for a town right on his doorstep.

This is the comfortable part of the story, where the new site beats established names and both Google and the AI agree he deserves it. The question is why. That is the part worth your time.

What we actually did to earn it: the E-E-A-T build

None of this happened because of a clever trick or a backlink campaign. There wasn't one. It happened because we built the site around the four things Google's quality systems and the AI assistants are both trying to measure: Experience, Expertise, Authoritativeness and Trustworthiness. Here is exactly what that looked like, pillar by pillar, on a real locksmith's website.

Experience. Every area page is written from real work, not a template. The Penkridge page does not say "we cover Penkridge and surrounding areas". It talks about grade-II-listed cottages and Victorian timber doors, about the 1970s uPVC estates, about dated call-outs from April and May, in the owner's own voice. One line reads: "Penkridge is a 10-minute drive from Brewood up the A449, and one of the closest towns I cover. It shares the ST19 postcode with my home village, so it's effectively part of my home patch." No national chain can write that sentence, because it isn't true for them.

Expertise. The site names the actual standards and parts a competent locksmith works with: BS3621 mortice locks for period properties, TS007 anti-snap cylinders for modern doors, gearbox brands like Lockmaster, Fullex, GU and Mila. One passage explains, in plain terms, that "the pocket dimensions on most Victorian doors fit a modern BS3621 mortice without any joinery work, so the door looks identical from the outside". That is the kind of detail you only have if you have actually stood at the door with the lock in your hand. It signals competence to a human reader and gives an AI something precise and verifiable to repeat.

Trustworthiness. The pricing is on the page, in public, while most of the trade hides everything behind a quote form. There is a dedicated, honest guide to how much a locksmith should cost that lays out real 2026 prices against typical UK ranges, explains the six things that move a price, and includes a scam-avoidance section with three real overcharge cases (£320 billed where £90 was fair, £480 for £145, £390 for £150). It says the quiet part out loud: "A locksmith who only drills is selling you a replacement you do not need." That page is the single biggest impression-gainer on the whole site, up 525% in Google Search Console, because it answers a question the whole industry avoids.

This is the pattern Google has been describing in its own words. Lily Ray, VP of SEO Strategy and Research at Amsive, summed up the 2026 shift cleanly:

"Google recently reminded everyone that 'non-commodity content' is what wins in search, meaning content with a unique perspective, real E-E-A-T, and genuine lived experience."

Lily Ray, Amsive (30 April 2026)

The first time I read that, the phrase that stuck was "genuine lived experience". You cannot fake it and you cannot scale it with AI. A one-man locksmith has it in abundance: every page can carry a real job, a real street, a real price he actually charged. We did not invent his expertise, we just made sure the site showed it. That is the same playbook we used on a removals firm in the same village, the one ChatGPT later scored 99 out of 100 against the national brands. Different trade, identical principle.

Authoritativeness. This is the one a small firm builds slowest, and it is where the story gets honest. His authority signals are strong locally: a 5.0 rating from 32 Google reviews, a consistent name, address and phone number everywhere, and the full suite of machine-readable AI Discovery Files at the root of the site so assistants can read exactly who he is. ChatGPT clearly did read them; asked about the business directly, it recited his coverage, his services and his real prices accurately, to the pound.

Thirty-two reviews is plenty to win a village. It is not plenty to win a city against a rival with thousands. Hold that thought, because it is the hinge of the next section.

Page one on Google, invisible on ChatGPT

Here is the result that makes this more than a victory lap. In Wolverhampton, Lockerfella ranks number five organically on Google, on page one, for "locksmith wolverhampton" and again for "emergency locksmith wolverhampton". That is a hard term to crack in two months.

Now run the same blind prompt through ChatGPT with search on, "Who's a good locksmith in Wolverhampton?", and he is nowhere. The model names city-branded rivals instead: Arvco, Wayne Lewis, JR Locksmiths, LockRite, LockFit. We ran the emergency variant too, same outcome. Page one on Google, absent from ChatGPT, for the identical search on the identical day.

This is not a glitch, and it is not the model being unfair. It is three things stacking up, and they are worth understanding because they apply to your business too. First, ChatGPT with search is not reading Google's live rankings; it reads its own sources and weighs them its own way, and it leans hard on review volume. It openly cited rivals with "216+" and "5,921" reviews against his thirty-two. In a small town that gap does not matter; in a contested city it is decisive. Second, brand-and-location identity: Arvco, JR and LockRite read as Wolverhampton-first businesses, while Lockerfella reads as Brewood and South-Staffordshire-first, with Wolverhampton as a covered area. The model picks the firms whose whole identity is the city. Third, the clocks are different. A Google ranking that climbed this month does not reach an AI assistant's sources on the same schedule, if it reaches them at all.

The cleanest proof that this is per-platform, not a flaw, is that Google's own AI does the opposite of ChatGPT here. As we documented when the May core update completed, Google's AI Overview already surfaces Lockerfella for a head term he doesn't even rank for organically, because it is built on Google's index where his content is strong. Same business, same day, three AI surfaces, three different verdicts. Mark McNeece, who created the AI Visibility Definition standard, frames why this happens in one line:

"AI Visibility is not about being found. It is about being understood."

Mark McNeece, AI Visibility Explained (April 2026)

That distinction is the answer to the Wolverhampton puzzle. Google has crawled and ranked his Wolverhampton page, so he is found. But ChatGPT, weighing the city's review-heavy incumbents, has not yet been given enough to understand him as a Wolverhampton choice specifically. As Mark puts it elsewhere on that page, "If AI is not sure who you are, it often just leaves you out." The fix is not an SEO trick. It is more of exactly what already won the home towns: more reviews tied to Wolverhampton, stronger Wolverhampton-specific content, more local citations. The gap is a to-do list, not a defeat.

Where the AI correctly says no

We deliberately tested a city well outside his core patch, to see whether ChatGPT would name him anyway. It didn't, and that is the most reassuring result in the whole exercise.

For Birmingham city centre, ChatGPT recommended J.F. Locksmiths and The Lock Surgeon, then described Lockerfella honestly: covers Birmingham, but based in Brewood with the city on his outer ring, "a good option if you want Sean's direct-owner service, but maybe not the most central pick for an immediate city-centre job". That is exactly how Sean would describe his own Birmingham coverage. A case study where the AI picks your client every single time should make you suspicious. One where it sensibly picks someone else, for a defensible reason, is one you can trust the rest of.

What this means if you run a small business

Strip out the locksmith specifics and three lessons travel to any local business.

First, a brand-new site can compete fast, in a hard niche, with no aged domain and no link-building budget, if the content is truly non-commodity. The long-tail and well-defined local searches come first; the biggest head terms take longer because they reward review volume and prominence you have to earn over time. Patience on the head terms is not failure, it is the normal shape of the curve.

Second, ranking on Google and being recommended by AI are different goals with different signals. You can be page one on Google, top of Google's AI Overview, and invisible on ChatGPT, all at once, for the same search. Chasing a single "number one" misses most of the board. Treat each surface as its own channel. If AI visibility is new to you, our sister service explains the foundations in AI Visibility Explained, and you can grade your own site free with the AI Visibility Checker.

Third, the thing that moved the needle on every surface was the same thing: specificity backed by accountability. Real prices, real streets, real standards, a named owner who stands behind it. That is what a well-built website is for now, and it is the opposite of the generic "fast, friendly, 24/7" pages most local sites still ship. Speed matters too, the site is hand-built for a perfect mobile PageSpeed score, but speed without substance ranks nothing.

One last honesty note, because it is the rule we hold ourselves to. Every position here is a dated snapshot from 21 June 2026, taken from one location, and local results shift by the day and by where the searcher is standing. We are not claiming a permanent leaderboard. We are claiming a clear pattern at one point in time, and we plan to re-test the whole set in August 2026 to see how it moves, especially whether ChatGPT catches up on Wolverhampton. You can hold us to that follow-up.

Frequently Asked Questions