What we actually do
Four disciplines, ordered by how well the evidence supports them. Two of them are mechanical and cheap. One of them is slow and carries most of the leverage. The fourth is how you find out whether any of it worked.
Short answer
An assistant names a business when it can retrieve a page that says the thing, and when enough independent sources corroborate it. Everything below follows from those two conditions. We do the mechanical half first because it is cheap and verifiable, then spend the rest of the effort on the half you do not own, which is where the evidence points.
Ordered by how well the evidence supports them
Not by how impressive they sound. The first is a mechanical prerequisite, the second is where most of the leverage sits, the third shapes the page for retrieval, and the fourth tells you whether any of it moved.
How an engagement runs
Three stages. We will tell you at the end of the first whether the rest is worth doing, and we have told people it is not.
01. Establish a baseline you can argue with
We run a fixed set of local-intent prompts for your category and region across ChatGPT, Claude, Perplexity and Google AI Overviews, repeatedly rather than once, and record what comes back. In parallel we fetch your site the way the crawlers do, with JavaScript disabled, and read your robots.txt and CDN rules.
You get the prompt-set results with the sample size and spread, a crawler access report with the exact lines quoted, and a list of the facts about your business that no source states clearly. That last one is usually the most uncomfortable document in the engagement.
02. Fix what demonstrably blocks retrieval
Content served from the server rather than assembled in the browser. OAI-SearchBot, Claude-SearchBot, PerplexityBot and Bingbot unblocked in both places they can be blocked. Dead URLs and redirect chains cleared. The paragraph under each H1 rewritten to answer the question the page exists for.
Schema goes in too, because it is cheap and it earns conventional rich results. Google states no special structured data is needed for its AI surfaces, and a matched study found adding JSON-LD produced no major uplift in citations anywhere. We deploy it and we describe it honestly.
03. Work on the half you do not own
This is the slow part and the part with the most leverage. Branded web mentions correlated with AI Overview visibility at 0.664 against 0.218 for backlinks across roughly 75,000 brands. Correlational, as the authors say and as we repeat. It is still the strongest published signal in the field.
We sample which sources the assistants actually cite for your category before targeting anything, then work on getting you into those. Then we re-sample monthly and fix whichever constrained query you lost.
This is a fit if
- You run an independent hospitality business where customers choose by asking rather than by browsing: a restaurant, a boutique hotel, a specialty café, a tour or experience operator.
- You can make changes to your own website, or you have someone who can. Most of stage two is unglamorous technical work.
- You are willing to have facts published about your business that you currently keep vague: exact hours, an honest price band, party-size limits, what you do not do.
- You can wait. There is no index to submit to and no switch to flip, and anyone promising a date is guessing.
It is not a fit if
- You want a guaranteed position. There is no position and no submission form, and the only Search Console control for AI features is an opt-OUT. Nobody can sell you one honestly.
- You want a monthly dashboard number that always goes up. Ours reports uncertainty, and in a flat month it will say the month was flat.
- Your site cannot change. If the content is assembled in the browser and that cannot be altered, the mechanical prerequisite fails and the rest is wasted money. We will say so on the first call rather than take the retainer.
What we will not sell you
We have published the full list, with sources. The short version: llms.txt does nothing for visibility and Google says so outright, speakable never left a news-publisher beta, FAQPage stopped earning a rich result on 7 May 2026, schema does not cause citations, and the widely quoted “40 percent uplift” is a maximum from a synthetic benchmark rather than a client outcome.
Three of those five were on this site until we audited our own copy in September 2026 and took them off. That page records which ones. Who we are and why we work this way, if that matters to you.
Four hospitality categories, and the queries that decide each
Common questions
A good deal of the underlying work is shared, and we will say so rather than invent a distinction to justify a rate. Crawlability, server-rendered HTML, clean site structure and fixing 404s serve both. Where it genuinely diverges: AI crawlers other than Googlebot and AppleBot do not execute JavaScript at all, an unlinked brand mention carries weight where SEO valued the link, and there is no ranking to report, so measurement has to be a sampled prompt set rather than a position.
It depends on how much of stage two your site needs, which we cannot know before fetching it. The first call is thirty minutes and free, and it ends with a specific answer about which prerequisites are failing rather than a proposal to think about.
We will give you dates for the implementation work, because that is ours to schedule. We will not give you a date for the answers changing, because there is no index status to check and no platform publishes its retrieval cycle. Treat any specific timeline promise, ours included, as a sales tactic.
Rarely, and we would rather tell you that than take the work badly. The method generalises but the value here is knowing which constraints decide a dining or accommodation query and which sources get cited for them, and that knowledge is category-specific.
No. There is no submission form and nothing you can file that puts you in an answer, so nobody can promise placement. What we can commit to is fixing the mechanical prerequisites, working the off-site half, and reporting a sampled prompt set honestly, including when it has not moved.
Start with a baseline
Book a 30 minute call. We will fetch your site the way the crawlers do, read your robots.txt live, and tell you which of the four prerequisites is failing. If none of them are, we will say so and you will have lost half an hour.