The sources AI trusts — and why your website isn't one of them.

Engines are built to discount self-description, for the obvious reason: every business website says the business is excellent. A model that weighted self-praise would recommend whoever wrote the most confident homepage. So the engines treat your site as a claim and third-party coverage as evidence — and compose answers from evidence.

This inverts where most businesses put their effort. The website gets redesigned every two years; the third-party record — the thing engines actually quote — grows by accident, if at all.

Run a Perplexity query for any service plus a major city and read the citations under the answer. Do it for twenty queries and the pattern is unmistakable: the same handful of sources, over and over. In most categories: the review platform that owns the category (G2, Clutch, Yelp), the directories the industry actually uses, a handful of city-level editorial outlets, and Google's own business data.

That short list is the actual battlefield of AEO. Not the whole internet — those specific publications, per category, per engine. Each engine's list differs slightly, which is why a business can be visible on Perplexity and absent from Gemini, and why we map citations per engine in every query map rather than assuming one list fits all.

Round-ups punch above their weight for a structural reason: a "best providers in [city]" listicle maps perfectly onto the three-names-per-answer format — the engine's extraction work is already done. One placement in a credible round-up routinely outperforms months of self-published content.

Each source type has its own playbook, and none of them are fast: review platforms reward sustained depth — service-level detail, Q&A activity, a live review cadence. Directories reward complete, consistent, claimed profiles. Editorial rewards an actual pitch: a writer who covers your category, a story worth telling, and proof you're worth quoting.

Citations also compound: editorial mentions feed the training-data layer, strengthen your knowledge panel, and make the next pitch easier. It's the slowest work in AEO and the most durable.

The advice above assumes you have something to build on. Plenty of excellent businesses don't — no editorial trail, a skeletal review-platform profile, nothing for an engine to quote. The cold-start sequence matters, because pitching national trade press with no supporting record is pushing rope.

First, the floor: claimed, complete, consistent profiles on every directory that matters in your market — they're weak signals individually, but they corroborate your entity and they're fully under your control. Second, review-platform depth, because it's the one authority you can build presence on without anyone's permission: service detail, Q&A participation, a review cadence. Third, local before national — a city-magazine quote or neighborhood round-up is a realistic first placement, and engines answering city-level queries lean on city-level sources anyway. National editorial comes easiest when a writer can find an existing trail; you're building the trail that makes the bigger pitch land.

Expect the cold start to take a season. The encouraging part: in most markets, almost nobody is doing this deliberately. The bar for "best-cited business in the market" is lower than the quality of work being done would suggest — which is exactly the gap AEO exists to exploit.

If you're allocating between another website refresh and a year of deliberate citation building, the citation work wins on every AI-search metric we track. Your site needs to be clean, fast, and machine-readable — schema included — and then it has done its job. The record that gets you named lives on other people's domains. Build there.