Data to AI™ / Gemini/YouTube
Give analysts and developers YouTube views, watch time, comments and playlist activity where they already work: the terminal, a notebook folder or the VS Code project of your docs site.
Better answers, fewer tokens.
Gemini CLI
Which traffic sources brought the most watch time to our September uploads?
Suggested videos drove 63% of watch time across these three sources, about twice as much as YouTube search, and held viewers longest at 46% average view duration.
| Traffic Source Type | Views | Watch Time (Minutes) | Average View Duration (%) |
|---|---|---|---|
| Suggested videos Owned Channel Videos Traffic Sources | 48,200 | 212,100 | 46% |
| YouTube search Owned Channel Videos Traffic Sources | 31,500 | 101,000 | 38% |
| External Owned Channel Videos Traffic Sources | 9,800 | 21,600 | 27% |
Views and Watch Time (Minutes) are summed per traffic source; Average View Duration (%) is weighted by Views, not a plain average of videos.
YouTube Organic extracts your channel's non-paid performance: views, watch time, subscribers, traffic sources, audience, comments and playlists, per video and per day.
Get Video ID, Title and View Count from Videos Metadata, then find every YouTube embed under docs/ that points to a video missing from that list.
A developer maintaining a product documentation site runs this in Gemini Code Assist before a docs release. Embeds of deleted or replaced tutorials show up as a list of files to fix, instead of as broken players that readers report weeks later.
Write Date, Video ID, Views, Watch Time (Minutes) and Subscribers Gained from Owned Channel Videos Basic User Activity for the last 90 days to data/youtube_daily.csv.
A video analyst keeps a Jupyter notebook that charts growth per upload. Gemini CLI regenerates its input file on demand, and the commit history shows which version of the data each chart was drawn from.
Find this week's comments whose Text Original reports an error or a broken link, and draft one markdown entry per Video ID in issues/youtube-feedback.md.
A developer relations team runs a channel of product tutorials. Gemini turns viewer complaints into a reviewable file in the repo, so fixes to code samples are tracked like any other issue.
Data to AI™ is priced per seat, like the AI assistants your team already uses, with ~25% off on annual billing. Need tailored contracts or volume pricing? Talk to us about Enterprise.
Serve governed business data to Claude, ChatGPT and other AI tools over MCP.
$10.49/month
per seat
3 seats minimum
Flexibly deploy Dataddo in any cloud or hybrid environment.
Custom
We have a payment model that works for you
| Without Dataddo |
With Dataddo
|
Outcome for you | |
|---|---|---|---|
| Scripted YouTube pulls | Every team that wants YouTube numbers in a script registers its own Google OAuth client and handles tokens inside each repository. | Google is authorized once in Dataddo by an account with admin-level permissions; Gemini CLI only signs in to Dataddo and reads the scheduled YouTube Organic flow. | One Google authorization |
| What an agent may run | A homemade script with broad API credentials runs whatever it was written to do, and nobody sees each request. | Gemini CLI shows each Dataddo tool call and waits for your confirmation; choose Always allow only for the tools you are comfortable with. | Each call approved |
| Viewer names in committed files | A comment dump written into a repository brings Author Display Name with it, and the names stay in git history. | Untick Author Display Name when refining the Comments attributes in Dataddo, so the files Gemini writes carry only comment text, likes and reply counts. | No viewer names in git |
| Token consumption | Gemini spends tokens exploring raw YouTube columns, sampling rows and retrying until it understands the data. | Field definitions and relations are supplied up front, so the model skips the exploration and goes straight to the answer. | Fewer tokens per answer |
| Answer quality | Prone to wrong joins, invented fields and numbers that are hard to check. | Answers grounded in defined fields. In the Dataddo benchmark, 78.5% of questions were answered correctly, against 65.5% with plain CSV files. | Better answers |
The YouTube datasets marketing teams use most, with their real field names. Gemini queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask Gemini
|
|---|---|---|
| Owned Channel Videos Basic User Activity |
|
Write Date, Video ID and Subscribers Gained for the last 28 days to a CSV that our growth notebook reads. |
| Owned Channel Videos Traffic Sources |
|
Which external sites in Traffic Source Detail send us viewers but are missing from the partner list in this repo? |
| Owned Channel Videos Demographics |
|
Which Country Code holds more than 10% Views Percentage on each video, as JSON for our subtitle script? |
| Owned Channel Videos End Screens Performance |
|
Which videos show zero End Screen Element Clicks, so I can check the end screen template in our upload tooling? |
| Videos Metadata |
|
Which videos have no Tags, listed by Title and Published At, so I can fix them in the upload manifest? |
| Comments |
|
Which comments in Text Original report an error or a broken link, grouped by Video ID for the issue tracker? |
| Playlist Basic User Activity |
|
Which Playlist ID values had no Playlist Starts at all, so I can archive them in our channel config? |
We asked 43 pre-registered questions of three real datasets - Google Search Console, HubSpot deals and Google Ads - and delivered the same rows to the same model in four ways. Only the data layer changed.
of questions answered correctly - the best of the four delivery methods tested
fewer confidently wrong answers - 14.0% of runs against 24.0% and 26.3%
more correct answers on messy CRM data - better on 8 questions, worse on none
Same model, prompts and rows in every condition; 759 graded runs, ground truth frozen before any run. On Google Search Console alone, documented CSV files did as well as Data to AI™ - the advantage comes from messier data such as HubSpot deals.
Read the benchmark →Create a YouTube Organic source in Dataddo with datasets such as Videos Metadata and Owned Channel Videos Basic User Activity, and attach it to a Google Gemini destination or AI Model. Add the httpUrl https://headless.dataddo.com/mcp-data under mcpServers in ~/.gemini/settings.json, restart Gemini CLI and run /mcp auth dataddo to sign in.
No. This setup runs through Dataddo: the YouTube Organic source extracts your channel data on a schedule, and Gemini CLI reads it from the Dataddo Data Access MCP server after you sign in with your Dataddo account.
Use ~/.gemini/settings.json to reach YouTube data from every project, or .gemini/settings.json in one repository to limit it there. In both cases use the httpUrl key, since url falls back to SSE.
No. Google and Microsoft SSO are not offered for this sign-in; use your Dataddo email and password, with MFA if prompted. Gemini CLI opens the sign-in page in a browser, so it needs one on the same machine.
Not in the source. YouTube Organic datasets do not use a date range, so each scheduled run pulls all currently available data. Limit the period in your question, for example by Date.
Yes. Agent mode in VS Code reads the settings.json that Gemini CLI uses. In IntelliJ, Gemini Code Assist has its own mcp.json, where you add the Dataddo server separately.
No. The Data Access MCP server is read-only over the extracted data. Gemini asks for entities and fields by name, Dataddo never accepts SQL from the model, and nothing is written back to your channel.
