Disclosure: I work on FORMAT, a tool that turns interior renders into short videos. The keyword data and Search Console numbers below come from our own accounts. No affiliate links.
I work on FORMAT, a small tool that turns one interior render into a 15-second video. Before writing a line of marketing copy, I did what every SEO guide tells you to do: keyword research.
The result was the most useful negative result I've had.
Every interior-specific keyword measured zero
I used Google Trends (US, 12 months) with gpts as a fixed benchmark, because absolute volume was not available to me. Relative to the benchmark:
| Phrase | Relative interest |
|---|---|
photo to video |
64.8 |
image to video ai |
58 |
gpts (benchmark) |
13.2 |
real estate video |
31.2, falling to ~4 by the time I checked |
room to video |
51.5 average, but collapsing week over week |
interior design video |
0 |
interior design video generator |
0 |
interior design video maker |
0 |
architectural visualization |
0 |
interior walkthrough |
0 |
3d render video ai |
0 |
Every phrase that described our actual customer returned zero. The only phrases with volume were generic: "photo to video", "image to video".
That felt wrong, so I checked a competitor. Lupa AI is a paid product in the same niche. Its brand search volume was also zero. If people were finding tools like ours through Google, a paid product with customers would have some branded search. It had none. Its search results page was seven TikTok, Instagram and YouTube links and two official pages.
The conclusion: for this category, demand is stimulated, not pre-existing. People see a video on social media, learn the category exists, then look for a tool. They do not wake up and Google "interior design video generator".
What we changed
We retargeted the main page from the natural-but-dead phrase ("AI interior video generator") to the phrase with actual demand ("photo to video"), while keeping "interior design" as the qualifier. Same visitor, same product, better-matched entry point.
We kept the use-case guides — render-to-video, Reels, client presentations — but stopped pretending they would bring search traffic. They exist to answer questions for people who already arrived, and to give AI answer engines something specific to quote.
The part nobody warns you about: AI answer engines
I asked ChatGPT, in a fresh session, two kinds of questions.
Branded: "What is FORMAT by VIXL?" It answered accurately, including the fixed five-move camera path and the 15-second length.
Category: "Best AI tool to turn an interior render into a short video?" FORMAT did not appear at all. It recommended Kling, Runway, Luma, Veo, Seedance, Pika and Firefly.
When I asked it to explain why, the answer was blunt: the information about FORMAT came almost entirely from FORMAT's own website, and that does not count as independent evidence. LLMs recommend what third parties recommend.
That changed our roadmap more than any keyword spreadsheet did. We stopped treating our own marketing pages as the main lever and started treating them as the thing a third party can point at.
The numbers after three months
Real Search Console numbers, not a growth screenshot:
- 93 impressions, 11 clicks in three months.
- Every query was branded (
vixl,vixl ai, and some misspellings). - Zero non-brand impressions on any FORMAT-related phrase.
- 9 of 53 pages indexed. Most of the rest were "discovered, not indexed".
None of that is alarming for a three-month-old site with no links. It is also not a strategy. Impressions follow indexing, indexing follows links and crawl demand, and we had neither. Publishing more pages would have added to a queue, not to traffic.
Five things I'd tell someone starting today
- Check that the demand exists before you check difficulty. I nearly built for a term with a low difficulty score and no searches. Low KD on a zero-volume phrase is a trap: it is easy to rank and worthless to rank.
- A competitor with zero branded search is information. It usually means distribution is not search. Look at where their traffic actually comes from.
- For stimulated demand, the product has to be visible where the stimulation happens. That means social, creators, communities — not a comparison table on your own domain.
- GEO is a third-party problem. Your own page can be accurate, complete, and still ignored by LLMs. Independent reviews, directories and community posts are what get quoted.
- Narrow tools win on trust, not features. Our best-performing content was honest about limits — fixed camera path, 15 seconds, possible drift. The pages that said "it just works" converted worse than the ones that told people when not to use it.
What I'm doing next
Directories and community mentions, in that order, with a strict rule: only where the product is actually relevant. Then a creator or two in the interior visualization niche, because that is where this category is discovered.
If you are building a tool for a professional niche and your keyword research comes back empty, I would not immediately conclude you picked the wrong niche. I would conclude that Google is the wrong front door — and go find the right one.
I work on FORMAT (vixl.art). No affiliate links. The keyword data and Search Console numbers are from our own account.
Top comments (1)
Great negative result, and the conclusion about stimulated demand rings true. One caveat worth adding: a 0 in Trends means 'too low to show relative to the other terms', not zero searches, and comparing against a big benchmark like gpts pushes small niche phrases to 0. Absolute monthly volume settles it. Google Keyword Planner shows ranges for free with an Ads account, and exact numbers need a paid source. (Disclosure: we built a bulk volume Actor on Apify for this, apify.com/steady-scrape/keyword-search-volume, but Keyword Planner is enough for a handful of phrases.)