Most of the advice circulating about AI search optimisation was written for companies with a content team, a PR retainer and a budget line for brand mentions. If you run a plumbing firm with four vans, or a domestic electrical business where the owner still goes out on jobs, almost none of it is actionable. Entity disambiguation across a multi-brand portfolio is not your problem. Co-occurrence programmes across tier one publications are not your problem.
Your problem is narrower and more fixable: when someone asks an AI assistant for an emergency plumber in your town, does the machine know you exist, know what you do, and trust the facts it has about you enough to pass them on.
That is the whole game at this scale. Everything below is about winning it.
Local service queries are a different species
When someone asks an answer engine something generic, say how a heat pump works, the system has a vast pool of acceptable sources. Any competent explainer will do. It can synthesise from several and nobody gets hurt if it blends them.
A local service query is not like that. "Find me a gas safe engineer in Crewe who can come out today" has a correct answer set that is small, geographically constrained, and carries real consequences if it is wrong. Sending someone to a business that closed last year, or one that does not cover their postcode, is a visible failure.
The practical effect is that answer engines handle these queries conservatively. They lean on sources that are structured, verifiable and recently confirmed, rather than on prose they have inferred meaning from. Mapping data, business listings, review platforms and your own site’s structured markup all carry more weight here than a well written paragraph on your homepage, because they are explicit rather than interpreted.
This is genuinely good news for a small business. You cannot outspend a national brand on content. You can absolutely be more accurate and more complete than your local competitors in the handful of places that matter, and accuracy is cheap.
The second effect worth understanding is that answers to local queries tend to be short lists rather than essays. There are usually three or four businesses named. There is no page two. You are either in the set or you are invisible, which makes the margin between "almost there" and "cited" much sharper than it was in classic search rankings.
Your Google Business Profile is the backbone
If you only do one thing after reading this, make it this one.
Google’s local results have long been built on Business Profile data, and AI generated answers about local services draw on that same local layer. When an answer engine needs to state that you open at eight, cover a particular area, and do boiler repairs as well as installations, the Business Profile is the most structured, most frequently verified statement of those facts that exists anywhere.
Completeness matters more than most owners assume, because an incomplete profile does not just look sparse. It creates uncertainty. A profile with no service list cannot confirm you do the thing being asked about. A profile with no service area cannot confirm you cover the postcode in the query. The system does not guess generously on your behalf.
Work through the whole thing properly:
Categories. One primary category that describes what you mainly do, then secondary categories only for things you genuinely offer. Resist the urge to stack ten categories to cast a wide net. Unrelated categories muddy the picture of what your business is.
Services. List them individually, in the language customers use. "Emergency boiler repair" is a service. "Plumbing" is a shrug. Each entry gives a query something to match against.
Service areas. Define the towns and areas you actually cover. If you are a service area business without a shopfront, this is the only thing establishing your geographic footprint.
Hours, including exceptions. If you offer out of hours call outs, make that explicit rather than hoping people infer it. Bank holiday hours matter more than they seem, because "today" queries are extremely common in trades.
Photos of real work. Not stock images of smiling models in clean overalls. Job photos with your van, your team, your finished work.
Reviews, and replies to them. Review text is a rich source of plain language detail about what you actually do and how you do it. Replying signals the profile is live and tended rather than abandoned.
The description and attributes. Write it as a factual summary, not a sales pitch. Say what you do, where, for whom, and since when.
None of this costs anything beyond an afternoon. Very few small trade businesses do it fully, which is precisely why it is worth doing.
Schema that actually matters on a trade site
Structured data advice online is usually a checklist of thirty schema types, most of which are irrelevant to you. Here is the short version of what a small service business site genuinely needs.
LocalBusiness markup on your homepage, or a more specific subtype if one fits. Schema.org has specific types such as Plumber and Electrician. If one matches your trade, use it rather than the generic LocalBusiness. The specific type tells a machine what you are without requiring it to interpret your copy.
Inside that markup, the fields that carry real weight are the ones that map to how people ask questions: your legal and trading name, address or service area, telephone number, opening hours, and the URL. Get these exactly right, and make sure they match your Business Profile and your directory listings character for character. A different formatting of your phone number across sources is not fatal, but a different number entirely is a problem.
Service markup on your individual service pages. If you have a page about emergency boiler repair, mark it up as a service with a defined area served and a link back to the provider (you). This is the connective tissue that lets a machine understand that this specific service is offered by this specific business in this specific place.
FAQPage markup on pages where you genuinely answer questions. Real questions, with real answers, of the kind you get asked on the phone. Not padded filler questions written to trigger a rich result.
Review or AggregateRating markup only where you have legitimate reviews to point at, and following the platform’s rules on self serving reviews. Faking this is both against guidelines and easily detected.
What you can skip: elaborate breadcrumb hierarchies on a nine page site, Product markup when you are selling labour rather than goods, and the long tail of experimental types that have no bearing on local service answers. More markup is not better markup. Accurate, minimal and consistent beats comprehensive and sloppy.
One practical warning. Many small sites have schema injected automatically by a theme or plugin, and it is frequently wrong: outdated addresses, placeholder opening hours, the developer’s test data. Check what is actually being output rather than assuming the plugin knows your business. Google’s Rich Results Test and the Schema Markup Validator will both show you what a machine sees, and both are free.
Why citations beat guest posting for a plumber
Here is where the enterprise advice does the most damage.
A citation, in the local sense, is simply a mention of your business name, address and phone number on another site: directories, trade bodies, review platforms, local chamber listings, your professional registration body. The value is not link equity. The value is corroboration. Multiple independent sources stating the same facts about your business makes those facts more credible to a system that has to decide whether to repeat them.
Inconsistency does the opposite. If four directories list four slightly different versions of your trading name, or two of them carry the mobile number you stopped using three years ago, you have introduced doubt. A machine resolving "which business is this and can I trust these details" now has conflicting evidence. In a context where the answer engine is conservative by design, doubt costs you the citation.
Compare that to a guest posting campaign. For an enterprise brand, earned mentions across respected publications build the kind of broad entity recognition that helps a model understand what a company is. For a local plumber, a guest post on a general industry blog does almost nothing for the query "plumber near me in Stockport", because that query is resolved through the local data layer, not through general web authority. You would be buying a solution to a problem you do not have.
Spend the equivalent effort on an audit instead. List every place your business details appear. Correct the ones that are wrong. Claim the ones you have never claimed. Add listings on the obvious platforms you are missing, especially trade specific ones and your registration body’s public directory, which carries genuine verification weight. It is tedious and unglamorous, and it is worth considerably more to you than a byline.
How to check what AI currently says about you, for free
You do not need an enterprise monitoring tool to do this. You need twenty minutes and a notepad.
Open ChatGPT, Gemini, Perplexity and a normal Google search in separate tabs. Use a logged out or private browsing session where you can, so you are not seeing results personalised to you.
Then run the questions your customers actually ask. Not "best plumber", which nobody types. Things like:
- "Who can fix a leaking radiator in [your town] today"
- "Recommend a gas safe registered engineer near [your postcode]"
- "Is [your business name] any good"
- "What does [your business name] do"
- "How much does a boiler replacement cost in [your area]"
Record three things for each. Were you mentioned at all. If so, were the details correct. And which sources were cited, where citations are shown.
That last column is the most useful output of the exercise. It tells you which platforms the answer engines are actually leaning on for your trade in your area. If three different systems keep citing the same two directories, those directories are now your priority for cleanup, regardless of what any generic list of top citation sites recommends.
Repeat it quarterly. Keep the notes. The pattern over time is more informative than any single snapshot, and it costs you nothing but the time.
What to ignore
Being direct about this saves you money.
Digital PR retainers. Built for brands that need national awareness. If your customer base lives within a thirty minute drive, this is spending on reach you cannot convert.
Multi-brand entity strategy. Relevant to organisations managing several brands that a model might confuse. You have one business with one name. There is nothing to disambiguate.
Co-occurrence and brand mention programmes. The theory is sound at scale. At your scale, the same budget spent on review generation and listing accuracy affects the local data layer directly, which is what actually resolves your queries.
Mass content production. Publishing forty generic articles a month will not get you cited for a local service query, because the answer engine has no reason to prefer your version of a generic explainer to a national publisher’s. Six genuinely local, genuinely useful pages will do more.
Anyone guaranteeing AI rankings. There is no ranking to buy, no placement to sell, and no mechanism that would make such a guarantee meaningful.
The order to do it in
Fix the Business Profile first, fully, in one sitting. Then run the free audit above so you know which sources matter for your trade and area. Then clean the citations those sources expose, starting with anything carrying wrong contact details. Then check and correct your schema. Then, and only then, think about content, and make it local and specific when you do.
That sequence works because each step makes the next one more effective. Clean data in the places machines read first, then build on top of it. For businesses that want this handled as an ongoing system rather than a one off project, Nimble Dingo’s AI growth systems are built around exactly this order of operations, and you are welcome to start with the free audit yourself and bring us whatever it turns up.