Home › Guides › The directories and citation surfaces AI trusts
Guide
The single most useful column in any AI-visibility audit is "who wins instead" — and right behind it, "where did the engine find them?" Because that's the to-do list. Firms don't get recommended by having a nice website; they get recommended by being present, consistently, on the handful of third-party surfaces the engines actually pull from. Here's the ranked map.
When an assistant builds a shortlist, it favors names it saw corroborated across independent sources. Your website is not independent — you wrote it. It clears the "does this firm exist" bar, but on its own it rarely clears the "should I stake a recommendation on this firm" bar. Independent corroboration is the currency, and these are the mints.
| Surface | Why engines lean on it | Effort to earn |
|---|---|---|
| Vertical directories & professional-body listings (e.g. NAPFA, a state bar directory, CPA society, specialty registries) | Curated, category-specific, and treated as semi-authoritative. High signal-to-noise for "best X for Y" questions. | Low–medium (apply, complete profile, keep it current) |
| Review platforms (Google Business Profile, industry-specific review sites) | Density and recency of reviews reads as consensus. Perplexity in particular leans here. | Medium (earn reviews over time; keep the profile accurate) |
| Independent editorial "best of" lists (local business press, niche publications) | Reads as third-party endorsement; a single inclusion can tip consensus. | Medium–high (earned, not bought; pitch or get noticed) |
| Community discussion (Reddit, forums, Q&A threads) | Reddit shows up in a large share of LLM citations — engines treat lived-experience threads as candid signal. | Hard to engineer honestly; earn it, don't astroturf |
| Your own website + llms.txt / schema | Necessary for grounding and to be found at all — but self-authored, so weak as a standalone signal. | Low (you control it) |
Note the shape: the surfaces that move answers most are the ones you control least. That's uncomfortable, but it's the honest picture, and it's why "add three more blog posts" is not a strategy.
The instinct after reading the table is "list everywhere." Resist it. Two failure modes are common:
The right move is targeted: find the two or three surfaces the winners in your category appear on, get on those, and make the details identical everywhere.
Generic directory lists are almost useless because the surfaces that matter are category- and locality-specific. The reliable way to find yours is to read them out of your own audit:
This is exactly what the diagnostic's action plan does at scale: it reads your specific losses — which questions, which engines, which competitors — and names the surfaces to get onto, in priority order, instead of handing you a generic checklist.
Paid directory profiles are fine where the directory is genuinely cited. But editorial inclusion and community credibility can't be honestly purchased, and engines increasingly discount sources that read as promotional. There's no shortcut around being a firm worth recommending and then making sure the record reflects it. The good news, from the transfer data: more than half of every local market's incumbents are currently invisible in this channel, so the firms that do this deliberately have an open lane.
The free scan emails you the "who wins instead" column for your category across the major engines — the first step to reverse-engineering your target list. The full diagnostic names the surfaces in priority order.
Get my free scan Full diagnostic — $490Related: Run the audit → · How assistants pick who to cite →