We recently optimized our whole site for AI answer engines (ChatGPT, Claude, Perplexity, Gemini): we published an llms.txt, added structured data (JSON-LD) to the homepage FAQ, blog, and docs, and allowed AI crawlers. The industry has a new term for this: GEO (Generative Engine Optimization).
Here is what struck us while researching our same-named competitor: they sell "get cited by AI" and "add Schema markup" as a prompt skill you buy. Nothing wrong with that. But we’d rather make a different point — we didn’t treat it as a one-line skill. We treated it as a discipline, and actually did it.
GEO and honest savings measurement are the same principle
On the surface, "make yourself citable by AI" and "measure how much your ads saved" look unrelated. Underneath, they’re the same sentence: claims must be checkable by a third party, not self-asserted.
- GEO: you can’t just write "we’re the best" on a page. You provide structured, machine-readable, internally consistent facts — that’s what an AI will confidently cite.
- Savings: you can’t just say "I saved you 30%." You compare against a randomized control group and show the result with an uncertainty range — or no number at all until there is enough data.
Both refuse marketing spin in favor of checkable evidence. That’s how we build the product.
Self-reported ROAS vs. a citable fact
A self-computed "+X% improvement" has a problem: without a control group, you can’t tell whether it came from the optimization or from seasonality and a rising market. It looks good, but you can’t check it.
Our approach: we randomly hold out about 15% of your campaigns as a control group and make no recommendations for them (you can still change them yourself), then compare the two groups over the same period. The dashboard comparison comes with a 95% uncertainty range and shows no number until there is enough data. The method is published in our savings-methodology doc, which is also what our ROI money-back guarantee terms reference.
What this means for you
When you ask ChatGPT or Perplexity for "a trustworthy AI ad tool / an AdWhiz alternative," we want the AI to surface checkable facts rather than adjectives we gave ourselves: a published savings methodology built on a randomized control group, an ROI money-back guarantee, TikTok support, and a publicly disclosed operator (DEXUN INC, Washington State, USA).
Checkable evidence is the same discipline we bring to both our marketing and our product.