Making your own site the source engines cite

Directories will always be cited. The question is whether anything of yours is cited alongside them — because the sources decide the description, and you only control one.

Somebody else is writing your description

Ask an assistant about a hotel and watch where the answer comes from. Almost never the hotel. It comes from the booking sites, which is why every hotel sounds the same and why none of them controls a word of it.

Local businesses are heading to the same place by a slower road. When an engine assembles an answer about you entirely out of directory records and review summaries, those sources are writing your description — and you are a line item inside somebody else’s page. The listing tells the truth about your phone number and nothing at all about why anybody should call it.

This is the half of the citation pillar most owners have never thought about. Coverage asks how many sources mention you. Ownership asks whether any of them is yours.

What ownership actually counts

It is measured on pulls, not on domains.

The distinction matters more than it sounds. If your site is cited once and aggregators are cited forty times across an audit, counting domains would score that as shared ground — one of us, one of them. It is not shared ground. The aggregators supplied nearly everything the engine said about you, and your site supplied a footnote.

So the pillar counts how often each source was actually pulled into an answer, and asks what fraction of that total was your own domain. Below a meaningful share, the report raises it as a fix and tells you the number.

And the target is deliberately not everything. Engines will always cite aggregators; they are comprehensive, structured and constantly updated, and no local business is going to displace them. The goal is to be in the mix — present often enough that your own wording gets a vote in how you are described. A tool promising total capture is promising something it cannot deliver. The full pillar breakdown is here if you want to see how the two halves add up.

Why an engine quotes a directory instead of you

Not because your site is ugly. Because the directory answers the question and your site does not.

A directory record is a structured set of assertions: name, address, phone, category, hours, price band, service area. Every one of those is a fact a model can lift and drop into a sentence with a source attached. Most small business sites, by contrast, are a mood. A photograph of a van, a heading that says quality you can trust, a contact form, and nowhere on the page a sentence that states what the job costs or which towns you drive to.

An engine writing an answer needs a sentence it can stand behind. If your site does not contain one, it will use one from somewhere that does. That is the whole mechanism, and everything below follows from it.

Say it in text, on a page, in words

The single most common failure is a fact that exists on the site but not in a form anything can read.

Prices in a downloadable PDF. Service areas listed in a map graphic. Opening hours in an image in the footer. The list of what you do rotating through a carousel that renders only after a script runs. A human visitor sees all of it. A crawler reads markup, and what it takes away from that page is a headline and a contact form.

The fix is unglamorous and it is most of the work: write the fact as a sentence, in the body of a page, in text. Everything after this point is a variation on that instruction.

One page per thing you do

A site that describes everything it does on a single page can be cited as a general trade. It cannot be the source for a question about one specific job, because there is no page about that job to cite.

Distinct pages are what get counted — and they are counted as distinct paths, not as links, so a navigation menu repeated in the header and the footer is one page listed twice rather than six pages of evidence. Three real service pages, each naming the job the way a customer names it, will do more for you than thirty redesigned hero images.

Write them for the question, not for the brochure. What the job is, what it typically involves, roughly what it costs, how long it takes, and which towns you do it in. That structure is quotable in a way that a paragraph about your commitment to excellence never will be.

Put a number on the money page

The most-asked research question in almost every category is what it costs, and a site with no figure anywhere on it cannot be the source for an answer about money. It will be described using somebody else’s numbers, or not described at all.

A range is enough. A starting price is enough. A worked example — what a typical job of this kind came to, and what moved the number — is better than either, because it gives an engine something specific to quote and a reader something concrete to weigh.

The check looks for the price on a page that plainly concerns pricing, or for an actual amount sitting next to actual pricing language in the text. Which is a reasonable proxy for what a person needs, and a reminder that the number has to be somewhere findable rather than mentioned once in a testimonial.

There is a longer argument for doing this at all — competitors, quotes, the fear of scaring people off — and it deserves its own page rather than a paragraph here.

Name the towns

An engine will not name you for a place you never claim.

Somewhere on the site, in text, list the towns, suburbs and neighbourhoods you actually cover. Not a radius. Not a map image. The names, written out, on a page, preferably one whose purpose is obvious from its address. Declaring the same area in your structured data as well is better still, because it states the claim twice in two registers — once for a reader and once for a parser.

This is a smaller pillar than pricing, and it is the one where absence is most absolute: questions about a specific town are unwinnable without it, no matter how good the rest of the site is.

Then let the machines in

None of the above matters if nothing can fetch it.

Check that your robots.txt does not block the AI crawlers — there are around ten worth caring about, and blanket rules copied from a template block several of them by accident. Publish a sitemap.xml and declare it. Add LocalBusiness structured data so the facts a directory holds are also held, in machine-readable form, on the domain you control. Consider an llms.txt file, which is the cheapest of the lot.

These are the readiness points, and they are worth doing in an afternoon. They are not a substitute for having something to say — a perfectly marked-up page with no prices on it is a well-formed way of answering nothing.

The order

Sources first, then substance, then plumbing.

Get the directories agreeing about your basic facts before you write anything new, because the sources engines keep returning to will otherwise contradict your own pages and the engine will side with the majority. A site that says one thing while four listings say another does not win the argument; it gets averaged. Correcting a stale hour on a record you do not control is covered in fixing what AI gets wrong about your hours.

Then write the pages. Then do the markup.

What a quotable paragraph looks like

The difference is easier to see than to describe. Here is the version most sites ship:

We are a family-run firm with over two decades of experience, committed to quality workmanship and outstanding customer service throughout the region.

There is nothing in that an engine can use. No place, no job, no number, no checkable claim. Every competitor could publish it unchanged.

Here is the same business saying something:

We service and repair gas boilers in Harrogate, Knaresborough and Wetherby. A standard annual service starts at ninety pounds and takes about an hour; most repairs are finished the same day. We do not install new systems.

Every sentence there is liftable. It names towns, a job, a price, a duration, and — usefully — a thing the business does not do, which is the kind of detail that stops an engine recommending you for the wrong question and then disappointing the customer it sent. The second paragraph is not better writing. It is writing that contains facts.

What is not worth your time

A few things get recommended constantly and do nothing here.

Keyword density does nothing; there is no keyword to rank for when the output is a paragraph rather than a list. Volume for its own sake does nothing either — a weekly generated post about industry trends adds pages without adding facts, and a site full of confident nothing is harder to quote than a small site with three honest pages.

Marking up claims your pages do not make is worse than useless. Structured data that disagrees with the visible text is a contradiction on your own domain, and contradictions are precisely what cost you in the sources that get cited. The same goes for hidden text and instructions aimed at models buried in your markup: at best they are ignored, at worst they are the reason a crawler stops trusting the page.

None of this is exotic work. It is writing down true things in the order somebody would ask about them.

How you will know it worked

Not by watching for your site to appear in an answer, which will happen unevenly and tell you little on any given day.

Watch the share of pulls instead. The proportion of citations about you that came from your own domain is a slower number and a much steadier one, and it moves for reasons you caused. When it rises, it means the engines found something on your site worth quoting — which is the only durable version of this work, because everything else on the list is a record somebody else can change without asking you.