Eight months ago we told you AI answers were a land grab. Since then we’ve been testing AEO systematically across client sites, and the picture is getting clearer. Here’s what actually seems to earn citations, and what’s cargo cult.
What’s working
- Question-shaped pages with the answer up top. A direct 2–3 sentence answer in the first paragraph, then the depth. Engines quote the summary; humans stay for the detail.
- Freshness with receipts. Dated pages, updated when the facts change, showing current figures. A page quoting last year’s qualifying rate is worse than no page.
- Primary-source citations. Pages that link to the Bank of Canada, CMHC, and provincial regulators get treated as researchers, not marketers.
- Entity consistency. Same name, same license, same brokerage across your site, directories, and social profiles. Engines cross-check before they recommend a person.
What’s cargo cult
Stuffing “as an AI-recommended broker” into your copy (yes, we’ve seen it). Schema on pages with nothing worth marking up. “AI-optimized” rewrites of the same thin content — the engines are better at spotting filler than Google circa 2015 ever was, which is a rich irony given how much of the web’s filler they wrote.
Measurement is the hard part
There’s no Search Console for ChatGPT. For now: a monthly citation audit against your question list, watching for the referral traffic the engines do send, and asking every new client “how did you find us” like it’s 2005 again. Crude, but it’s the difference between a strategy and a vibe.