Google has published an open-source tool called YouTube Topic Insights to GitHub, offering advertisers and marketing teams an automated way to identify trending content and top creators on YouTube without manual research. The project, housed under the google-marketing-solutions GitHub organisation, combines the YouTube Data API with Google's Gemini AI models to surface trends, popular videos, and channel-level intelligence through a Looker Studio dashboard.

The repository was created on January 22, 2026, according to the GitHub metadata attached to the project. The most recent code push occurred on February 4, 2026, when contributor letiziab-git submitted a commit described as "Fix prompts by reming old language references." Sara Marques, a Partner Engagement Manager at Google UK, shared the project on LinkedIn approximately three weeks before the time of writing, framing it as a tool built to solve a specific operational problem for advertisers.

"Ready to automate how you discover trends and top creators on YouTube?" Marques wrote in her LinkedIn post. "The tool leverages the YouTube Data API alongside Gemini's advanced content understanding to transform raw video data into actionable intelligence."

The announcement sits within a broader context of Google pushing Gemini capabilities into its marketing toolset. Google's NewFront 2026 presentation in late March positioned Gemini as the engine running across its entire marketing platform, while Google Marketing Live 2025 introduced a range of AI-powered features including open-source measurement tools. YouTube Topic Insights follows a similar pattern - open-source, AI-powered, and designed to reduce manual workflow burdens for marketing professionals.

What the tool does

According to the project documentation, YouTube Topic Insights is an automated Looker Studio dashboard that pulls data from the YouTube Data API and then passes it through Gemini's content and language understanding models. The pipeline works in four stages. First, the system queries the YouTube Data API for the most-viewed videos published within a configurable time window - the default example given in the documentation is the past 30 days - for a set of keywords defined by the user. Second, Gemini analyzes the retrieved video content, detects the language of each video, and generates a summary that identifies key elements and topics. Third, these AI-generated summaries are aggregated alongside performance metrics. Fourth, the combined output is surfaced in a centralized Looker Studio dashboard that presents three categories of intelligence: trending topics, top videos, and top creators.

The project states that "staying current on trending topics and content can be a manual and time-consuming task" and that the tool is designed to give advertisers "a fast and automated way to identify content that drives engagement."

Architecture and technical components

The system is built on two main components. The first is a Google Cloud Platform infrastructure layer managed using Terraform, an infrastructure-as-code tool that automates the provisioning of cloud resources. According to the README documentation, this component enables the required APIs - Vertex AI and YouTube Data API - and configures the permissions necessary for the script to operate. The second component is a Google Apps Script application that functions as the orchestrator, living within a Google Sheet that also serves as the user interface.

The Apps Script side consists of 11 distinct files. The main workflow file is appsscript.gs. Supporting files handle discrete functions: youtube_api.gs manages all interactions with the YouTube Data API; gemini_batch_module.gs handles communication with Vertex AI Batch Prediction; channel_analytics.gs contains the logic for analyzing YouTube channels; gemini.gs manages real-time calls to the Gemini API; prompts.gs holds all primary prompts sent to Gemini; automation.gs contains the self-triggering logic for scheduled tasks; and logger_module.gs handles debugging output to a dedicated sheet.

The use of Vertex AI Batch Prediction is notable. According to the documentation, this design choice allows the system to analyze hundreds of videos efficiently without hitting execution timeouts that would otherwise interrupt the process. The system uses what the documentation describes as a "chained-trigger" architecture, meaning tasks are processed sequentially and the automation chain is built to be "immune to global execution timeouts and capable of handling a virtually unlimited number of automated tasks." A single analysis task can time out if it encounters an exceptionally large number of videos - the Apps Script environment imposes execution limits of 6 minutes for a single run and 30 minutes in certain contexts - but the batch infrastructure is designed to handle this.

The entire GCP setup is defined and reproducible via Terraform. Key Terraform files include main.tf for defining cloud resources, variables.tf for declaring input variables, terraform.tfvars for user-specific values, and outputs.tf for defining what information Terraform surfaces after deployment.

The codebase is written primarily in JavaScript (94.7%) with the remainder in HCL (5.3%), which is the language used for Terraform configuration files, according to the GitHub repository metadata.

Configuration and search parameters

All search parameters are configured inside a Google Sheet rather than in code. The documentation specifies that the system can handle multiple complex search queries configured through a spreadsheet, set to run automatically on a schedule. Users enter values for country code, keyword, lookback days, and result sheet name into a Config sheet. The Automation sheet allows users to maintain a list of keywords, each with an enabled/disabled flag, so the system can cycle through a queue of topics on a recurring basis.

The intelligence the system develops about channels goes beyond videos. According to the technical documentation, the channel analytics component "goes beyond videos to analyze the channels behind the trends, summarizing their content and style" - a feature described as in-depth channel insights. This channel-level intelligence is relevant for advertisers evaluating potential creator partnerships, since it surfaces information not just about individual videos but about the creators producing them consistently.

The default Gemini model specified in the codebase is gemini-2.5-flash, defined in gemini.gs. The documentation notes that users can update this value to use different Gemini model versions as they become available, making the system forward-compatible with future model releases.

Language detection as a filter

One feature that stands out for international marketing teams is automatic language detection. According to the README, the system "automatically detects the language of each video and can filter out results that don't match your target language." This means a user configuring the tool for a specific market can restrict results to videos in a particular language, reducing the noise that would otherwise appear when a keyword draws results across multiple language communities on YouTube.

This matters in practice because YouTube operates at a genuinely global scale. A keyword search for a consumer category might return videos in English, Spanish, Portuguese, German, Korean, and Japanese simultaneously, with no native filtering applied by the YouTube Data API itself. The language detection layer built into YouTube Topic Insights addresses this by applying Gemini's language understanding to each video's content and filtering the results accordingly.

Requirements and setup

The tool requires four infrastructure prerequisites, according to the project documentation: Google Cloud with billing enabled, Google Workspace, the Gemini API enabled, and the YouTube Data API enabled. Beyond these, users need a Google API key, which can be generated through Google AI Studio by creating a new key associated with the relevant Google Cloud project.

Two setup paths are available. The quick start option involves copying a pre-configured Google Sheet template and linking it to the user's GCP project. The manual setup option requires copying the full codebase into a new Apps Script project file by file. Both paths ultimately connect the same Apps Script environment to the same GCP infrastructure.

Importantly, the project describes itself as requiring no coding to operate once set up, since all configuration happens through the Google Sheet interface. However, the setup process itself - deploying Terraform, cloning the repository, configuring a GCP project, setting OAuth consent screens - does require familiarity with cloud infrastructure tooling. The documentation recommends using Google Cloud Shell, which runs in a browser and comes with Terraform and the gcloud command-line tool pre-installed, specifically to lower this barrier.

Development history

The commit history reveals that the project was not always called YouTube Topic Insights. The earliest substantive commit, dated December 19, 2025, initializes what was then called "TrendTube," described as a project incorporating Apps Script, Terraform, and documentation. A January 12, 2026 commit updated the README with a new Google Sheet link and configuration guides. On January 22, 2026 - the same date the repository was made public - a commit by contributor FranceLando carried the message "refactor: rename TrendTube to YouTube Topic Insights." Two commits on January 23, 2026 added Copybara configuration for syncing with an internal Google codebase repository. The most recent commit, on February 4, 2026, fixed prompts by removing old language references - a likely consequence of the renaming from TrendTube to YouTube Topic Insights.

The repository carries an Apache License 2.0, permitting commercial use, modification, and distribution. It currently has 5 stars and 3 forks on GitHub. Two contributors are listed: FranceLando (Francesco Landolina) and letiziab-git. The project carries a disclaimer that it is not an official Google product.

Why this matters for marketing professionals

The tool arrives at a specific moment in YouTube's development as an advertising platform. YouTube closed its native trending page on July 21, 2025, removing a centralized surface that had existed since 2015 and that advertisers and media planners had sometimes used as a proxy for cultural relevance on the platform. YouTube's own explanation at the time was that trends had fragmented into numerous micro-communities rather than a single ranked list. YouTube Topic Insights directly addresses that gap by allowing advertisers to define their own keyword-based lens on what is trending, rather than relying on a platform-curated list.

The creator identification angle is also relevant. YouTube unified its creator partnership tools into a single platform called YouTube Creator Partnerships on March 24, 2026, launching in seven markets including the US, UK, Brazil, and India. A March 2026 Think with Google report found that 79% of Gen Z viewers trust YouTube creator recommendations, and that creator-led campaigns showed 2.3 times higher long-term return on ad spend compared with paid social. Against that backdrop, a tool that automatically surfaces top creators by topic category gives media teams a starting point for outreach that would otherwise require manual channel-by-channel research.

Google's open-source Meridian marketing mix model, released globally in early 2025, established a pattern of Google releasing analytical tooling as open-source software for the marketing community. YouTube Topic Insights follows that same pattern, applying it to content intelligence rather than media measurement. The earlier release of an open-source Google Ads API Model Context Protocol server in October 2025 added another data point to this trajectory.

YouTube's enhanced viewer analytics introduced in July 2025 - which segmented audiences into new, casual, and regular viewer categories - gave creators and brands more granular data about engagement depth. YouTube Topic Insights operates at a different layer, focusing on discovery and content intelligence rather than audience classification, but both tools reflect the same underlying direction: replacing manual research processes with structured, automated data pipelines.

For advertisers planning YouTube campaigns, the ability to scan a topic area for the most-viewed content published within a defined window, identify the creators responsible for that content, and understand the thematic structure of what is performing - all through a system that can be scheduled to run weekly - reduces the research overhead that has historically made systematic YouTube content intelligence difficult to sustain at scale.

Timeline

  • May 21, 2025 - gPS GoB automation adds initial README.md file to the repository on GitHub, establishing the project's earliest record in version control.
  • December 19, 2025 - Contributor FranceLando initializes the project under the name "TrendTube" with Apps Script, Terraform infrastructure, and documentation. (YouTube closed its trending page on July 21, 2025, removing the platform's native trend surface weeks before development begins in earnest.)
  • January 12, 2026 - README updated with a new Google Sheet template link and configuration guides.
  • January 22, 2026 - Repository made public under the google-marketing-solutions GitHub organisation. The project is renamed from "TrendTube" to "YouTube Topic Insights" in the same commit.
  • January 23, 2026 - Copybara configuration added for syncing with Google's internal codebase repository.
  • February 4, 2026 - Final code update: prompts corrected to remove old language references tied to the TrendTube name.
  • March 2026 (approx.) - Sara Marques, Partner Engagement Manager at Google UK, publishes the tool on LinkedIn, directing followers to the GitHub repository.
  • March 24, 2026 - YouTube unifies BrandConnect and the Creator Partnerships Hub into YouTube Creator Partnerships, launching in seven markets - contextualizing why creator identification tools are gaining attention at this moment.
  • March 29, 2026 - YouTube opens its Shopping affiliate program to creators with as few as 500 subscribers across 12 countries, further expanding the creator ecosystem that tools like YouTube Topic Insights are designed to help navigate.

Summary

Who: Google Marketing Solutions, a division within Google, developed and published YouTube Topic Insights. The project was built primarily by contributors FranceLando (Francesco Landolina) and letiziab-git, and is publicly promoted by Google UK's Partner Engagement team.

What: YouTube Topic Insights is an open-source, automated tool that combines the YouTube Data API with Google's Gemini AI models to identify trending topics, top-performing videos, and leading creators within any keyword-defined category. The output is delivered through a Looker Studio dashboard, managed through a Google Sheet, and deployed on Google Cloud Platform using Terraform. The system supports scheduled automation, language filtering, channel-level analysis, and batch video processing through Vertex AI.

When: The project was initialized internally in December 2025 under the name TrendTube. The GitHub repository was created and made public on January 22, 2026. The last code update was committed on February 4, 2026. Public promotion via LinkedIn began approximately three weeks before this article was written.

Where: The tool is available at github.com/google-marketing-solutions/youtube-topic-insights under an Apache License 2.0. It runs on Google Cloud Platform, using Google Workspace, the Gemini API, and the YouTube Data API as its operational infrastructure.

Why: Identifying trending content and top YouTube creators has historically required manual, time-consuming research. YouTube's own trending page was shut down in July 2025, removing a centralized discovery surface. YouTube Topic Insights automates the research process, applying AI analysis to public YouTube data and delivering structured intelligence to advertisers and media planners who need to make content strategy and creator partnership decisions at scale.

Share this article
The link has been copied!