Services · Video Engine Optimization (VEO)
YouTube SEO, rebuilt as VEO: one video for Google, AI answers and YouTube
In some sectors, up to 53% of search results show a video box your text pages can never win. AI assistants cite video transcripts as sources. Our published VEO framework builds one video that takes the box, feeds AI answers and ranks on YouTube. No camera, no production budget, sometimes in under two weeks.
What is Video Engine Optimization (VEO)?
Video Engine Optimization (VEO) is the practice of building one video so it performs on every engine at once: the Google video box, YouTube search, Google's AI Overviews and the AI assistants that cite video transcripts. Botanik published the full VEO framework as an open guide, workflow and checklist included.
VEO is not YouTube SEO. YouTube SEO optimizes for one platform. VEO makes one video work everywhere: it captures the Google video box, it gets cited by AI assistants through its transcript, and it ranks inside YouTube search. One production, three engines.
The framework is public, free tools included. This page applies it as a service: Botanik has industrialized the whole chain, from query scoring to measurement, and specializes in it. In-house, the workflow takes about fifteen hours of work per video. You keep the result, not the workload.
One thing this page will not do: promise view counts. VEO is a search discipline, not an audience play. The result that matters is the surface captured, query by query. For the measurement method and the pricing grid, go to the category pages: /en/generative-engine-optimization-agency and /en/oracle-method. Here, we stay on video.
Why video moved ahead, quietly
The share varies by sector: 37% on automotive queries in France, up to 63% on automotive queries in the US. The US SERP is your future. Every one of those boxes is space your text pages cannot take.
No source is cited more, and its citations grew 34% in six months. When Google's AI answers reach for video, they reach for YouTube.
AI engines cite YouTube 50 times more than TikTok. To them, YouTube is a library of transcribed, structured answers, not a social feed.
YouTube is also the world's second search engine. Publish one video and you compete on two search engines with a single production.
Between 37% and 53% of Google SERPs now display a video box, depending on the sector. French automotive queries sit at 37%; the same queries in the US reach up to 63%. The direction is clear: the US SERP is where your market is heading. Four numbers set the stakes.
The five surfaces one video can hold
The SERP block your text SEO can never win, present on up to 53% of SERPs in some sectors. Whoever holds it takes that share of the page from every text competitor.
The world's second search engine, with 41 million monthly French users. The same optimized video competes there with no extra production.
Video is the number 1 source of Google's AI answers. A structured, transcribed video is exactly the kind of source they select.
Assistants read text. A clean transcript turns your video into citable text, so the video itself becomes a source in AI answers.
Traffic without a query, driven by engagement signals. The same video covers this fifth surface with no extra production.
A well-built video works on 5 surfaces at once: the Google video box, YouTube search, Google's AI Overviews, the assistants ChatGPT, Perplexity and Claude, and Google Discover. The first of those is a surface your text pages can never take. One production covers all five.
The VEO workflow in 7 steps
The published VEO workflow runs in 7 steps, from business scoring to industrialization, about fifteen hours of work per video when run in-house. Every step below sits in the open framework, and the framework ships with 3 free tools you can use without an account.
- 01Business scoringThe 20 queries that actually drive your revenue. We start from business value, not from a keyword list.Deliverable : A shortlist of 20 revenue queries.
- 02SERP analysisFor each query: does the SERP already show a video box? If yes, there is a space to take before anything gets produced.Deliverable : A map of queries with a winnable video box.
- 03Script from the content that already ranksThe page that ranks is the source of the script. Video and page speak to the same intent, with the same semantics, so the video inherits the page's relevance.Deliverable : One script per target query, mirroring the ranking page.
- 04ProductionAI text-to-speech or filmed footage, both work. You can also recycle: a LinkedIn post becomes a Short, an existing video gets repositioned on another query.Deliverable : One long-form video, produced or recycled.
- 05YouTube optimizationShort title carrying the keyword, chaptered description, clean hand-made SRT transcript, tags, thumbnail.Deliverable : A fully optimized YouTube listing.
- 06On-site synergyEmbed on the target page, VideoObject structured data, internal links, one dedicated page per video, a dedicated video sitemap. This is what helps Google surface the video box.Deliverable : Markup, pages and video sitemap live on your site.
- 07Measurement and industrializationTrack AI pickups and referral traffic per query, then replicate the process on every winnable query.Deliverable : A per-query tracking sheet and a repeatable process.
How does a video get cited in AI answers?
Through its transcript. YouTube is the number 1 cited source in AI Overviews, up 34% in six months. Assistants read the text attached to your video: the SRT file, the chapters, the description. A video without a clean transcript is invisible to them.
Upload the SRT transcript by hand. Never rely on the auto-generated one: auto captions are noise, and the transcript is exactly what AI engines read. This single habit separates videos that get cited from videos that just play.
Do not count on Shorts for AI visibility. A Short carries no exploitable transcript; to an AI engine, a Short is TikTok. AI engines cite YouTube 50 times more than TikTok precisely because YouTube gives them transcribed, structured material to quote. The long-form video is the citable asset. A Short can be cut from it, not replace it.
On your site, VideoObject JSON-LD markup tells Google and the AI engines what the video contains and where it lives. That markup, plus one dedicated page per video and a declared video sitemap, is what turns a clip into a source.
How fast can a video reach the Google video box?
Under two weeks, in the case we publish. Five videos out of five reached first place in the video box of their target query in less than two weeks after publication, with €0 of filming budget. That speed is not a promise; it is what happened, under the conditions below.
The case: a national network of car dealerships. We started from pages that already ranked, on financing, trade-in and used-car trust queries. The videos were produced entirely with AI text-to-speech. The voice-over sounded a little robotic at first; we iterated.
Why it worked: each video was anchored on a ranking page with the same intent, carried proper VideoObject structured data and clean chaptering, and entered SERPs where a video box already existed. Result: 5 videos published, 5 first places, under 2 weeks, €0 of filming budget.
What we do not promise: view counts, subscriber growth, or a fixed timeline on every query. The video box depends on the query, which is why the workflow starts by testing each SERP before anything gets produced.
The VEO checklist that makes the result
Seven rules decide whether a video captures its surfaces. All 7 come from the published framework, and they are the difference between a video that ranks and a video that merely exists. Keep them next to every publication; each one maps to a step of the workflow.
| Rule | Why it matters | Périmètre réaliste | Horizon de résultat |
|---|---|---|---|
| Anchor each video on a page that already ranks | Same intent on the video and the page. The video inherits the page's relevance instead of starting from zero. | ||
| VideoObject structured data on the page that embeds the video | This is the markup that helps Google surface the video box and helps AI engines cite the video. | ||
| Title under 50 characters with the keyword, chaptered description of 250 to 1,000 words | Short titles survive truncation; chapters give engines a map of the content. | ||
| SRT transcript uploaded by hand, never the auto-generated one | The transcript is what AI engines read. Auto captions are noise. | ||
| One dedicated page per video | Google indexes only one video per page. Two videos on a page means one invisible video. | ||
| A dedicated video sitemap, declared in the main sitemap and robots.txt | It tells Google exactly where every video lives, without waiting for a crawl to find them. | ||
| Avoid the Short as your AI asset | No exploitable transcript. To an AI engine, a Short is TikTok, and AI engines cite YouTube 50 times more than TikTok. |
Three free VEO tools, no account needed
Paste a keyword, get a verdict: does Google already show video on this query? Tested on the French market, mobile SERP. If the box is there, there is a space to take.
Paste a YouTube URL and get the JSON-LD VideoObject markup ready to paste into your page. This is the markup that helps Google surface the video box and helps AI engines cite the video.
Paste your videos and the pages that embed them: the tool generates a sitemap-video.xml ready to submit to Search Console.
The framework ships with 3 free tools, usable without an account: a SERP video box tester, a VideoObject JSON-LD generator and a video sitemap generator. They cover the SERP analysis and on-site synergy steps of the workflow. Use them before you talk to anyone, including us.
Where VEO sits in Botanik's method
VEO is one discipline inside a measured system. Your visibility in AI answers is scored on three states, Absent, Mentioned and Recommended, under the published Oracle method. The method, the engines and the pricing grid live on their own pages; this page stays on video.
The measurement behind every Botanik engagement is the Oracle method: corpus, engines, scoring, thresholds and limits, published in full and versioned. Read the specification at /en/oracle-method before you sign with anyone, including us.
For what a program costs and how engagements are built, see /en/what-a-geo-agency-costs and the category page /en/generative-engine-optimization-agency. Reporting on video work uses the same three states as everything else: Absent, Mentioned, Recommended.
Related reading in English: /en/aeo-vs-geo-vs-seo for the vocabulary, /en/answer-engine-optimization-agency for answer engines, and /en/ai-seo-agency and /en/llm-seo for the wider AI search practice.
What does YouTube SEO cost at Botanik?
€890 per day (excl. VAT), the single day rate for every Botanik service, with an audit from €2,670 one-shot (Graine, the diagnostic tier). Botanik sells multi-day packages, not single days, and higher intensity never raises the day rate. The full grid, the 60-day audit deduction and 2026 market ranges are at /en/what-a-geo-agency-costs.
Do I need a camera or a studio?
No. In the dealership case, €0 went to filming: the videos were produced with AI text-to-speech and reached first place in their video box in under two weeks. Filmed footage works too. The ranking comes from structure, anchoring and markup, not from production value.
Is a Short enough for AI visibility?
No. AI engines cite YouTube 50 times more than TikTok because they read transcripts, and a Short has no exploitable transcript. To an AI engine, a Short is TikTok. Build the long-form video; that is the citable asset. A Short can be cut from it for distribution, not replace it.
How is VEO different from YouTube SEO?
YouTube SEO works on 1 platform: YouTube. VEO builds the video so it works on five surfaces at once: the Google video box, YouTube search, AI Overviews, assistant citations through the transcript, and Google Discover. One production, three engines. The full comparison is in the published framework.
Which AI engines can cite a video?
In Google's AI Overviews, video is the number 1 cited source, up 34% in six months. ChatGPT, Perplexity and Claude cite well-transcribed videos as text sources. In every case the engine reads the transcript and the markup, not the pixels, which is why the checklist above decides the outcome.
How do you measure the result?
Three states, Absent, Mentioned, Recommended, plus per-query tracking of the video box, AI pickups and referral traffic. The states and their thresholds are defined in the published, versioned Oracle method (/en/oracle-method), so you can read the scoring and its limits before you sign anything.
Case studies, measured results
How Pennylane strengthened its SEO presence on key accounting and financial queries
With Botanik, Pennylane structured a comprehensive SEO strategy combining technical optimization, educational content, and semantic targeting to strengthen its position in the financial management tools ecosystem.
Franck Neuenschwander · founder, author of the Oracle method
The method you just read, Organic Inference™, is mine: published, versioned, with its limits in writing. Fourteen years on the advertiser side, and the only agency I have ever worked at is the one I run: if you want to discuss it, you discuss it with me, not with a salesperson.
Thirty minutes with a consultant
Not a salesperson. We tell you which lever fits your case, and if none does, we say that too.
Are you absent, mentioned or recommended?