When an AI engine cites a business in an answer, it often doesn't paraphrase — it lifts a sentence nearly word-for-word. And the sentence it lifts, surprisingly often, is the meta description.
In our test runs we watched one company's meta description get rewritten into AI answers three separate times, almost unchanged. It read, in essence: "[Country]'s [category] platform. 500+ qualified vendors across 77 provinces."
Compare that with a description that never got quoted despite the site being technically flawless: "Send photos and symptoms on LINE, get a quote, book a time…" — a process walkthrough. Nothing in it answers "what is this, where does it operate, how big is it."
The shape that gets quoted
A quotable meta description is one sentence that could stand alone as an answer, built from three parts:
- Category noun — what you are, in the words a customer would use. "Property management platform." "Thai restaurant." "Logistics provider." Not "your trusted partner in excellence."
- Service area — where. A city, a country, "nationwide." AI engines answering local-intent questions need this to know you're relevant.
- A verifiable number — scale (customers served, vendors, branches), coverage (cities, provinces), speed (response time), or price. One is enough. It must be true and checkable — engines increasingly cross-reference, and so do humans.
Keep it under about 160 characters. Longer descriptions get truncated in previews and are less likely to be lifted whole.
Examples of the rewrite
Before: "We are passionate about helping businesses grow through innovative digital solutions tailored to your unique needs."
→ After: "Digital marketing agency in Bangkok: 120+ SME clients since 2019, campaigns in Thai and English."
Before: "Welcome to our website. We offer a wide range of quality services at competitive prices."
→ After: "Air-conditioning repair covering all 50 Bangkok districts, technician on site within 24 hours, from ฿500."
Every element in the "after" versions is a claim the business can prove. That's the discipline: if you can't verify a number, leave it out rather than invent it. A fabricated figure that gets quoted by an AI engine is a public correction waiting to happen.
Why this works
The engine's job is to answer "which one, where, is it credible?" A description already shaped like that answer is the lowest-effort, lowest-risk sentence for it to reuse. You are, quite literally, drafting the engine's sentence for it.
One caveat we always attach: doing this correctly makes you quotable, not guaranteed to be quoted. Which sources an engine picks varies by question, language and engine. What this fix does is make sure that when your page is on the table, there's a sentence worth taking.