“How do I get AI to mention my brand?” is a question we hear every week. Instead of answering in theory, here is one test we ran end to end: taking a product that AI assistants never recommended for a specific need and getting it into the answers.
The setup
The product was a SaaS platform for AI video translation and dubbing, in a category with several large players. Across AI assistants it was known as a general tool, but it was never recommended for one specific job: dubbing recorded church sermons into other languages.
We chose this niche on purpose. Interviews with customers showed a real need, but keyword tools showed almost no search volume. Live translation at services already had providers; dubbing recorded sermons had no clear leader. That made it a clean test: we could see whether the method worked without competing against an entrenched answer.
Before March 2026, no source connected the product with religious organisations except one customer story. ChatGPT did not recommend it for this task.
What we did
1. Tracked one prompt across every platform.
“What AI tool can dub our church sermons from English into Spanish and Korean?” We tracked it in ChatGPT, Google AI Overviews, Google AI Mode and Gemini, recording mentions, position and cited sources.
2. Built one page around how customers describe the need.
Not around keywords: around the words people used in interviews. What mattered to them was keeping the pastor’s voice, handling several speakers and full localisation rather than subtitles.
3. Added proof on our own site.
Customer stories from religious organisations, so the claim was backed by examples.
4. Published on external sites with one consistent narrative.
Several articles on third-party publications described the same angle: full localisation of sermons, keeping the original voice, multiple speakers. Each compared the product honestly with alternatives built for other jobs.
The results
After 30 days, for that one prompt:
| Platform | Share of answers mentioning the product | Average position |
| Google AI Overviews | 90% | 1.5 |
| Gemini | 73% | 3.8 |
| Google AI Mode | 63% | 2.7 |
| ChatGPT | 53% | 2.9 |
Across all models over seven days: 65.7% mention rate, average position 3, with more than 23 other brands appearing in the same answers.
What surprised us
Our own page mattered most for Google. In Google’s AI products, the page we built was the most-cited source for the recommendation. Owned content is not dead; it needs to answer the exact question.
Not every external placement got cited. Only two of the external articles appeared among cited sources. Publication choice matters more than publication count. In hindsight we would have checked first whether each site already appeared in AI citations for related topics.
ChatGPT behaved differently. Most of its sources were vendor websites that did not mention the product at all, and our own site was barely among them. It still recommended the product, so we track its answers by text, not only by citations.
A small mistake cost us. External articles linked to a near-duplicate page instead of the one we built, splitting the signal. One clear page per need, and every link pointing to it.
How to repeat this for your brand
- Pick a narrow need where you genuinely win. Not “best video tool”, but one situation where your product is the right answer.
- Track a few prompts on every platform before you change anything.
- Learn how customers describe the need from interviews, support tickets and communities.
- Build one page for that need, with proof.
- Get mentioned on external publications using the same narrative, choosing sites that AI systems can read and already cite. A marketplace such as brandmentions.io, which lists editorial publishers with their AI crawler access, DR, traffic and price, can shorten this step. Check citations yourself too.
- Measure, then move to the next need.
What this is not
One prompt in one niche is not a guarantee for every category. A niche without a leader is easier than a crowded one. But the mechanism is the same: assistants recommend what the sources they read describe clearly and consistently.


