Pulling the Levers

A named gap points to a fix. This page takes each gap from Reading the gaps and maps it to the tactic that closes it. Measuring and diagnosing are becoming things anyone can buy. The mapping below, especially for local businesses, is the part that is hard to copy, so it is where the work pays off.

Two things to fix first, whatever your gap

Before anything else.

Make sure the AI bots can read your pages. If GPTBot, PerplexityBot, Google's crawlers, and the rest can't get in, nothing else matters. Sites that lean heavily on JavaScript or block the crawlers are invisible in every citation study. Check this first.

For a local business, get the Google Business Profile and reviews in order. AI answers lean heavily on the Business Profile for local questions: categories, hours, attributes, photos, the Q&A. And reviews are content the model reads, so ask happy customers to name the specific service and the city or neighborhood, something like "fixed our flooded basement overnight in [their city]." A neglected profile is a hole under everything else.

If the model doesn't know you

Get into the sources models learn from and pull from.

• Real pages for each service and place you serve, that answer what a buyer actually asks: the local specifics, what a job runs, what to expect. Not thin templates.
• Get listed on the directories and publishers that get cited for your category. The citation map tells you which.
• Editorial and PR coverage: best-of lists, local news, trade and neighborhood publications. The engines lean toward publisher sites.
• A YouTube channel that answers real customer questions, with the question in the title. A cold channel with no audience has moved AI visibility inside a week.

If it knows you but won't pick you (the ghosting case)

You are already known, so this is about why it keeps choosing other people.

• Make the content and citations behind the recommendation stronger and more specific than whatever is getting named now.
• Get onto the local best-of lists and into editorial coverage, which carry real weight in who gets recommended.
• Earn real recommendations on Reddit and Nextdoor, which tip the scales between local providers.

If the gap is only in one channel

• Strong with web off, weak with it on. Work the live layer. Clean, short URLs, a tight answer near the top of each page, and ranking in regular search too, since pages that rank for more terms get cited more.
• Weak with web off, strong with it on. Build presence that sticks, so you are known without a search. Keep your name and details consistent everywhere they appear, and keep up steady editorial and YouTube presence so you end up in training over time.

If you're thin or missing on a service

• If nobody owns it, publish the page: real cost ranges, the questions buyers actually ask, the answer up top. This is the winnable one.
• If a competitor owns it, it is harder and slower. Either out-do them on depth for that exact service, or go after a more winnable angle of it, and chase coverage for that service specifically rather than your brand in general.

If a competitor gets credit for your differentiator

Usually the cheapest lever on this page, because the differentiator already exists. The fix is purely about making it findable.

• Give it a page of its own that says it plainly, in crawlable text, phrased the way buyers ask: "open Saturdays," "sedation for nervous patients," not marketing copy that talks around it.
• Put it everywhere the engines look: the Business Profile fields and attributes, the directories from your citation map, and structured data (hours especially, since engines tabulate those directly off pages).
• Make sure your own site agrees with itself. Conflicting hours or details across your pages give the model a reason to skip you and quote someone whose facts line up.

If you get recommended but not cited

This is usually the highest-value fix, because you already have the recommendation and just don't own the source behind it. Make your own site the thing worth quoting.

• Publish real prices and ranges. If you don't, the model makes up a range from forums and competitors.
• Answer the high-intent local questions plainly: insurance and coverage, what drives the price, what to expect.
• Localized pages with real depth about the city and the neighborhoods you serve, on clean URLs, with answers a model can lift straight off the page.

Use the citation map to know where to show up

The diagnose step already turned up who is getting cited for your category. That list is your target: get listed on those directories, earn a spot in those publications, and study the competitor pages that get cited to see what kind of page wins. If a competitor's service-and-city page is the cited source, that is the page you are missing.

What doesn't work

Worth being straight about the dead ends, because they eat real time.

• Doorway city and service pages cranked out at scale, footer zip-stuffing, and "call for pricing."
• llms.txt for visibility. Fine for governance, no measurable lift.
• Alt text and filenames when the facts aren't also in the body copy.
• Hidden text, markdown copies of your pages, and metadata-only tweaks.
• Monitoring on its own. A dashboard that doesn't lead to a change is just decoration.

Then keep an eye on it

Levers take time to land, so the last step is watching. The Tracking section shows the AI layer reacting: bots re-crawling the pages you changed, and referrals coming in as the recommendations follow.