Google's Ads Developer Relations team has published four episodes of Ads DevCast between March and May 2026, building a podcast series aimed squarely at the technical users who build and maintain advertising integrations. The episodes span 18 to 41 minutes each and are available on the Google Advertising and Measurement Developers YouTube channel. Taken together, they trace a clear line from strategic framing about AI in advertising all the way down to the operational mechanics of keeping API pipelines running without breaking.

The host across all four episodes is Cory Liseno, a developer relations engineer at Google who has been with the company since 2018. The series is explicitly for builders - developers, technical marketers, and engineers - rather than for campaign strategists. According to Liseno in Episode 1, a separate podcast called Ads Decoded hosted by Ginny Marvin covers the product strategy and campaign side; Ads DevCast is "specifically for the builders."

Episode 1: MCPs, Agents, and Ads. Oh My! (March 18, 2026)

The first episode, published on March 18, 2026, and running 18 minutes and 15 seconds, introduces the podcast and frames what Liseno describes as the "agentic shift" in advertising. It gathered 617 views in its first weeks, the highest view count among the four episodes at the time of writing.

The episode introduces three organizing ideas: being smarter, faster, and more accessible. Smarter refers to leveraging large language models to synthesize advertising data and produce actionable analysis. Faster is about reducing the time developers spend reading documentation, debugging queries, and iterating on API integrations. Accessible reflects a shift in who can operate as a technical user of Google's advertising products - not just trained engineers, but marketing power users who can now work directly with tools like MCP servers.

MCP servers take center stage as the primary technical subject. According to Liseno, an MCP server is "a service that facilitates your agent, such as Gemini CLI, to access data." The specific examples covered are the Google Ads MCP server and the Google Analytics MCP server, both of which allow agents to retrieve live advertising and analytics data through natural language prompts rather than requiring custom code.

The practical example involves a colleague, Matt Landers, who used the Analytics MCP server to ask an agent to generate a marketing plan. The agent pulled data from Analytics, generated campaign outlines, produced budget projections, and attached the reports it used to justify its recommendations. Liseno frames this as a reduction in the middle 40 to 60 percent of a typical workflow - the data gathering and analysis phase that normally requires manual effort - while the human still sets context at the start and reviews output at the end.

The other tool covered in Episode 1 is the Google Ads API Developer Assistant, a Gemini CLI extension that helps developers generate, debug, and run code against the Google Ads API using natural language. Liseno shares a real support case where a developer was struggling to report on the policy reasons why a keyword or ad was not approved. Pasting the developer's original support message directly into Gemini CLI with the assistant loaded, Liseno describes how the tool resolved the issue within minutes - identifying why each of the developer's previous query attempts had failed, writing the correct query, and running it directly in the CLI. The Developer Assistant reached version 2.0 in February 2026, adding agentic conversion diagnostics and full support for Google Ads API v23.

On accessibility, Liseno notes that the podcast deliberately uses the phrase "Ads technical community" rather than "Ads developer community" - a shift reflecting the fact that marketing professionals are now operating MCP servers and running direct API queries without going through the traditional software development lifecycle.

The episode closed with a community segment drawing on questions from Discord and LinkedIn. One question - "Do I need to actually be a developer to use your MCP servers?" - received a candid answer: there is some upfront technical setup involved, but the servers themselves are "a natural language to data processing platform." Non-developers can use them.

Episode 2: Many to 1 - Audience Management in Data Manager API (April 2, 2026)

The second episode was published on April 2, 2026, runs 24 minutes and 7 seconds, and had accumulated 347 views at time of writing. The timing was deliberate: it landed one day after April 1, 2026, the date on which Google blocked all new Customer Match uploads through the Google Ads API. As PPC Land reported in March 2026, that deadline was a hard stop - not a soft deprecation.

The episode features a guest, Melissa Ng, product manager for the Data Manager API at Google, in conversation with Liseno. It is, as PPC Land covered, as much a migration briefing as a product introduction.

The core argument for the Data Manager API is consolidation. Before it existed, managing audiences and conversions across Google's products required separate integrations for Google Ads API, Display and Video 360 API, and Google Analytics - each with its own schema, field definitions, and authentication flows. According to Ng, one large advertiser told Google they were spending over $1 million per year on API maintenance alone. The Data Manager API provides a single ingestion endpoint. A developer sends data once and it fans out to multiple downstream destinations - Google Ads, DV360, Campaign Manager 360, and Google Analytics - all through one unified schema. The API was launched on December 9, 2025.

The episode also documents a case study from Treasure Data, described as one of the first data partners to integrate with the API. According to Ng, Treasure Data reported 2x faster advertiser onboarding and an 80% reduction in engineering effort after migrating to the consolidated API. Treasure Data works across thousands of advertisers, making those efficiency numbers consequential at scale.

Confidential matching receives extended treatment. The Data Manager API implements it out of the box for any customer match upload. According to Ng, confidential computing is a type of privacy-enhancing technology (PET) that runs in a trusted execution environment - an isolated hardware partition where even the administrator cannot view the data being processed. When hashed or encrypted PII arrives at Google, only that environment decrypts it and matches it against Google's ID space. The matched IDs are returned; the raw PII never leaves the secure environment. The customer receives technical attestation that the process worked as designed.

Migration timelines are spelled out in the episode. Data partners have been notified to complete migration by March 2027. The DV360 audience ingestion API faces a separate shutdown also set for 2027. New Customer Match implementations must use Data Manager API immediately - there is no onboarding path via the older Google Ads API. The sunset date for existing Customer Match users in the Google Ads API was TBD at the time of recording, but Ng indicated it would be announced soon with ample lead time.

On the roadmap, Ng described plans for the rest of 2026: store sales support for Google Ads, event conversion support in Google Marketing Platform, and app support in Google Analytics. Data Manager API v1.6, released May 7, 2026, delivered both store sales ingestion and expanded Analytics event support, matching the roadmap Ng described in the episode.

A new diagnostics layer in the UI - the first time API stats would be visible directly in Google Ads - was also flagged as forthcoming. The Data Manager UI is accessible under the Tools section in the Google Ads interface.

Episode 3: Improving your Advertising Workflows (April 16, 2026)

The third episode, published April 16, 2026, is the longest of the series at 40 minutes and 39 seconds. It features a second guest, Matt Landers, also from Google's Ads Developer Relations team. Landers was identified in Episode 1 as having been one of the earliest internal advocates for MCP servers within the team. The episode had 196 views at time of writing.

The episode is structured as a practical walkthrough of tools available to both technical marketers and developers - "not just talking about the hype of AI but instead the utility of it," as Liseno frames it.

The first substantial section returns to MCP server workflows, extending beyond the individual-query examples from Episode 1 into multi-server combinations. One scenario presented involves diagnosing why a campaign with high click-through rate is generating few conversions. Historically this requires separate API calls to Google Ads for ad performance data and to Analytics for landing page behavior, then manually joining those datasets on Google Click IDs and aligning attribution windows. With both the Google Ads MCP and the Google Analytics MCP loaded in the same agent context, the same analysis runs in a single prompt session. Liseno provides a sample prompt: "verify the engagement rate for my landing pages, get landing pages with a high click-through rate, check their engagement rate and bounce rate, and if the engagement rate is low or the bounce rate is high, suggest some improvements." The agent compares ad copy against landing page content, cross-references Analytics engagement metrics, and returns a diagnosis - in the episode's example, identifying a 90% bounce rate caused by a content mismatch between the ad and the page it was pointing to.

A second workflow covers improving Smart Bidding data. Smart Bidding depends on the quality and completeness of conversion signals. Using the Analytics MCP server, a developer can identify high-intent user behaviors - such as spending more than two minutes on a pricing page - that are not currently being tracked as goals and not being sent to Google Ads. Adding those behaviors as secondary conversion goals provides the Smart Bidding algorithm with information it does not otherwise have.

Landers covers the Google Analytics Report Builder Sheets add-on, which received a major update that integrated Gemini. Previously, building a report required manually specifying dimensions, metrics, and time ranges through a complex form interface. With the update, a developer or marketer can type a plain-language question - "what's my highest converting landing page?" - and the add-on generates the corresponding API request, runs it, and imports the results directly into Google Sheets.

The AdMob SDK migration skill is introduced as an example of domain-specific AI tooling for the publisher side. Landers describes a skill - a context file that can be pulled into a coding agent - designed to help developers migrate between AdMob SDK versions. The skill is available in Google's documentation and can be referenced directly by any AI coding agent; it is not restricted to Gemini CLI.

The discussion of developer workflows includes Landers sharing that he had not written a line of code by hand in over six months. He frames domain knowledge as still essential - not for writing syntax, but for knowing what to ask for and how to validate the output. His example: he knew to suggest switching from a relational database to an OLAP database when optimizing certain queries, a decision the AI tool confirmed but would not have initiated.

The community segment addresses questions from Discord and LinkedIn. One, from Sarah on LinkedIn, asks whether natural language querying via MCP will eventually replace the need to learn a product's query language. Landers' response is nuanced: syntax has become less critical to memorize, but understanding the architectural constraints and limitations of the underlying systems remains important. "The more that you know about the domain that you're working on, the better output you're going to get from your tooling."

Episode 4: Managing your Integrations at Scale (May 7, 2026)

The fourth and most recent episode was published on May 7, 2026, runs 19 minutes and 50 seconds, and had 93 views at time of writing. It accompanies a post on the Google Ads Developer Blog that summarizes the episode's key takeaways. There is no guest in this episode - Liseno delivers it as a structured technical walkthrough.

The subject is operational discipline at high volume. As Liseno sets up at the start, the episode is not about integrations pushing a few hundred keywords or uploading weekly conversion reports. It is about systems handling "thousands, tens of thousands, even millions of operational requests flowing seamlessly through automated pipelines without continuous oversight, without maintenance, without triage."

Four universal principles are framed as applying across every Google advertising API.

The first is rate limits. Every API imposes them, and according to Liseno, hitting them is "inevitable" at scale - the goal is to hit them as seldom as possible and to handle them gracefully when they do occur. Error messages often contain explicit guidance: some will specify a retry interval. The recommended pattern is exponential backoff with jitter - waiting 2 seconds, then 4, then 8 - rather than rapid retries that hammer an endpoint and risk extending the cooling period. Concurrent requests to other accounts can continue while one account's limit is being respected, and operations should be grouped to reduce total request count.

The second principle is logging and monitoring. Official client libraries typically include built-in logging that can write to local disk as a baseline. More practically, Google's support teams request replicable logs when developers file issues. According to Liseno, having those logs implemented from the start means "you can just copy and paste them directly into the thread and move on to solutioning." The associated Google Cloud project for any enabled API provides a record of all API calls, including version usage - useful for planning upgrades.

The third principle is using official client libraries where available. The libraries handle pagination automatically, manage authentication and token refresh, and provide default timeout handling. Developers who roll their own wrappers for supported languages miss those defaults and introduce surface area for bugs.

The fourth principle is caching. Data that Google will no longer update - historical campaign performance prior to a certain date, for instance - can be stored locally and served from that local copy. This reduces API calls for read-heavy workloads and can improve response times for end users. Recent data continues to require live API calls.

Product-specific guidance covers five surfaces:

For the Google Ads API, the key technique is batching operations. Rather than sending one mutate request per operation, developers should group related operations - for example, 1,000 campaign updates in a single call rather than 1,000 separate calls. Ordering matters: if a request contains campaign operations and ad group operations, they should be grouped by type. Interleaving them - campaign, ad group, campaign, ad group - causes the backend to make twice the number of internal service calls, degrading performance significantly. Automated assets in Performance Max and AI Max reduce the effort of maintaining creative at scale, but Liseno notes that programmatically applying brand guidelines can enforce stylistic exclusions across large campaign sets.

For the Data Manager API, the distinction between synchronous and asynchronous errors is critical. Synchronous errors - malformed JSON, missing required fields - are returned immediately in the HTTP response and reject the entire batch. If one of 1,000 objects in an array has a structural error, all 1,000 must be resubmitted. Asynchronous errors are discovered after ingestion, during internal processing, and are returned via the status endpoint. Developers need separate handling logic for each category.

For Google Ads scripts, sequential iteration across accounts is described as "the enemy of speed." The executeInParallelmethod in manager account scripts launches up to 50 execution threads simultaneously and extends maximum execution time from 30 minutes to an hour. Bulk CSV uploads are recommended over iterative object mutations in loops.

For the Display and Video 360 API, YouTube line items, ad groups, and embedded targets cannot be modified directly through the API. The workaround is Structured Data Files (SDFs) - bulk edit formats that can be downloaded via the API, modified programmatically, and either uploaded back to the UI manually or processed through automated pipelines.

For the Google Analytics API, large reports can time out under standard HTTP connection constraints. The solution is long-running asynchronous report tasks, available in the v1 alpha version of the API. Rather than keeping the connection open while data is filtered, a developer dispatches the report payload, receives an immutable task ID, and polls for completion using a background worker. When the status returns as succeeded, the completed dataset is available for download in segments.

Liseno closes with a three-question framework for any integration design: is it efficient, is it resilient, and is it auditable?

Context for marketing professionals

The four episodes collectively map a period of significant change in how Google has structured access to its advertising infrastructure. The Ads MCP server was released open-source on October 7, 2025, operating in read-only mode for reporting and diagnostics. The Google Analytics MCP server followed in July 2025. In February 2026, Google faced a developer token application backlog, citing the Developer Assistant, the MCP server, and the new Explorer Access tier as the drivers of unusually high application volumes. And in April 2026, Google unified all advertising developer resources into a single hub, housing the Ads DevCast playlist alongside API documentation and Discord links.

The podcast has accumulated 1,253 total views across its four episodes as of writing, with Episode 1 holding the highest count at 617. The channel running it has 11,300 subscribers. That is a modest audience in platform terms, but the series targets a specific segment - developers and technical marketers actively building integrations - for whom the content has direct operational relevance.

For marketing teams without deep engineering resources, Episodes 1 and 3 are most immediately applicable: the MCP server walkthroughs demonstrate workflows that can be adopted without building new software. For development teams managing large-scale advertising pipelines, Episode 4 is the most operationally dense. Episode 2 remains time-sensitive for any team still managing Customer Match or DV360 audience data through legacy APIs, given the deadlines Ng outlined.

Timeline

  • March 18, 2026 - Ads DevCast Episode 1 published: "MCPs, Agents, and Ads. Oh My!" - 617 views; covers Google Ads MCP server, Google Analytics MCP server, and Google Ads API Developer Assistant
  • April 1, 2026 - Google blocks new Customer Match uploads via Google Ads API; all new implementations must use Data Manager API
  • April 2, 2026 - Ads DevCast Episode 2 published: "Many to 1: Audience Management in Data Manager API" - 347 views; features Melissa Ng (Data Manager API PM) on migration timelines, confidential matching, and the Treasure Data case study (PPC Land coverage)
  • April 6, 2026 - Google launches unified Advertising and Measurement Developers Hub, housing Ads DevCast, Discord, and API documentation under one domain
  • April 16, 2026 - Ads DevCast Episode 3 published: "Improving your Advertising Workflows" - 196 views; features Matt Landers on multi-MCP workflows, the Analytics Sheets add-on, AdMob SDK migration skill, and developer AI tooling
  • May 7, 2026 - Ads DevCast Episode 4 published: "Managing your Integrations at Scale" - 93 views; covers rate limits, logging, caching, batching, DV360 SDFs, and async report tasks across five Google advertising APIs
  • May 7, 2026 - Data Manager API v1.6 released with store sales ingestion for Google Ads and expanded Analytics event support - matching the roadmap Melissa Ng described in Episode 2
  • March 2027 - Deadline for data partners to complete migration from legacy APIs to Data Manager API; DV360 audience ingestion via legacy API also scheduled to end

Summary

Who: Google's Ads Developer Relations team, led by host Cory Liseno with guests Melissa Ng (Data Manager API PM) and Matt Landers (Developer Relations Engineer). The series targets developers, technical marketers, and engineers working with Google's advertising and analytics APIs.

What: Ads DevCast is a biweekly video podcast covering technical advertising topics across four episodes: the agentic shift and MCP servers (E1), Data Manager API migration (E2), AI-driven developer and marketing workflows (E3), and large-scale integration architecture (E4). The four episodes together document a range of tools including the Google Ads MCP server, Google Analytics MCP server, Data Manager API, Google Ads API Developer Assistant, Analytics Sheets add-on, AdMob SDK skill, and best practices for the Google Ads API, DV360 API, Google Ads scripts, and Analytics Data API.

When: Episodes were published on March 18, April 2, April 16, and May 7, 2026.

Where: The podcast is available on the Google Advertising and Measurement Developers YouTube channel and as an audio podcast. Resources, surveys, and Discord links are listed in each episode's description.

Why: The series reflects Google's effort to equip the technical community - developers and increasingly non-developer technical users - to work more effectively with its advertising APIs at a moment when AI tooling is changing both what is possible and who can do it. Migration pressures, including the April 1 Customer Match deadline and the March 2027 data partner sunset, give several episodes immediate practical urgency beyond general education.

Share this article
The link has been copied!