Starting point
Turn the blog posts into scripts, in Polycam's voice.
Polycam makes a 3D scanning app. Their product is used by architects, engineers, and contractors who use phone scans for renovation documents, material estimates, and scan-to-CAD workflows. Each blog article they published spoke to core concerns of these users and our brief was to turn them into short videos that retained their voice while optimizing Answer Engine citations.
Polycam's blog post-to-video pipeline had two gaps we had to solve. Its drafts were written for viewers, so they opened with context and buried the answer an AI would lift out. And some claims had no third-party source.
What we found
A video that speaks to non-human viewers.
Answer engine optimization, or AEO, is creating content that an AI system answering a question quotes. For a video, it reads the description, chapter markers, transcript, and, as far as we can tell, most reliably a dense closing summary. This structure turned out to also suit human viewers even though it disrupted the standard flow: the answer came first, the hook second.
Where we went
Eight rules, and a fact basis stapled to every script.
We researched best practices, industry standards, ran a competitor analysis, and developed a playbook for the pipeline. The rules we came up with: answer the title's question in the first sentence. Make each section a question and answer that works when quoted alone. Source every number. Close on a dense summary written for extraction, not storytelling. Keep the call to action out of that summary. Be honest: disclose Polycam's limitations, because AI engines read that as trustworthy. Match on-screen text to the spoken claim. Put the story of the problem after the answer.
Six videos were scripted, shot, edited, and launched using this process. The first four covered a material takeoff from a phone scan, a scan-to-CAD DWG workflow, a comparison of five floor plan apps, and generative versus detection-based AI. Each script came with a fact-basis appendix: every claim about another company traced to its own pricing page or documentation, with the URL and date checked.
The comparison video was the hardest and required us to develop a new research methodology. We compared five companies' products, but each one makes different outputs, uses different pricing models, and needs different devices. So we priced and specced one hypothetical test case for each product, at live prices, and held Polycam to the same standard. The script offered new, useful knowledge about the whole 3D scanning industry. Polycam later searched YouTube's floor plan videos, theirs and everyone else's, for how often answer engines cite each one. This one came back highest.
Citations do not follow views.
About 1,600 people watched the floor plan comparison in its first two months. Answer engines cited it 77 times.
What's next
The eight rules are now a system prompt.
Each new video starts from that prompt and a source article. Polycam reported the citation rate doubling in the month after the first videos went up, and the floor plan comparison drawing 77 citations on its own. Both are their measurements, not ours, and the citation counts come from an AI-assisted search rather than a meter, so we treat them as early signals, not settled results. The engagement is ongoing.