Google Ads MCP and 2026 Updates: What Media Buyers Actually Need to Know

i manage $2.4m in annual ad spend across google and meta for lead-gen businesses. when google shipped an official mcp server, force-upgraded dsa to ai max, moved to monthly api releases, and sunset the content api in the same 12-month window, i had to rebuild my operating assumptions. this post covers what changed, what it means for lead-gen buyers, and what to do about it.

six updates defined google ads in 2026. first, the google ads mcp server shipped in april, giving ai agents read-only access to account data without custom integration code. second, ai max moved out of beta in april with a september force-upgrade deadline for dsa, aca, and campaign-level broad match. third, the google ads api shifted from quarterly to monthly releases in january, compressing the upgrade window from 12 weeks to 4. fourth, performance max gained first-party audience exclusions in march, letting buyers block existing customers from prospecting campaigns. fifth, ask advisor launched as google’s cross-product ai agent in may, connecting ads, analytics, and merchant center in one surface. sixth, ai brief gave media buyers natural language control over ai max behavior, and the content api sunset on august 18, forcing shopping feed migrations. here is what each one means for your cpa and lead volume next tuesday.

google ads mcp: the official read-only server for ai agents

google released the official google ads mcp server on april 28, 2026. the repo, github.com/googleads/google-ads-mcp, has 914 stars and 270 forks as of september 2026. it is apache 2.0 licensed, python-based, pipx-installable, and uses stdio transport. before mcp, connecting an ai agent to google ads data required custom api integration code, oauth handling, and gaql query construction. now an agent can query your account in plain language.

the mcp server exposes two core tools: list_accessible_customers and search. the search tool runs google ads query language (gaql) against your account. companion tools include get_resource_metadata, discovery-document, metrics, segments, and release-notes. setup requires a google developer token, a google cloud project with the ads api enabled, and oauth 2.0 credentials or a service account. the setup is heavier than meta’s one-click oauth flow, per soku.ai and digital applied. but once configured, your ai agent has structured access to campaign performance, keyword data, and conversion metrics without manual csv exports.

the critical limitation: google ads mcp is strictly read-only. it cannot modify bids, pause campaigns, or create assets. per google’s own documentation and soku.ai’s verification, the read-only boundary is intentional. google’s safety model prevents accidental spend changes through ai agents. this is a reporting and analysis integration, not a buying tool.

for a deeper dive on how i use ai agents (including mcp integrations) for campaign research and optimization, see my guide on ai agents for media buying.

ai max force-upgrade: dsa, broad match, and aca are dying in september

google announced in april 2026 that ai max is moving out of beta. ai max is not a new campaign type. it is an optimization layer inside existing search campaigns with three engines: search term matching, text customization, and final url expansion. google claims 7% more conversions at similar cpa/roas when all three features are enabled. independent testing tells a more nuanced story.

the force-upgrade timeline is locked in:

• august 3, 2026: creation of new campaign-level broad match and aca blocked
• september 1-30, 2026: auto-upgrade of existing broad match and aca campaigns
• february 2027: dsa auto-upgrade begins (delayed from original september 2026 timeline)
• february 2027: dsa creation permanently blocked
• september 2026: in-account notifications start appearing

dsa was the catch-all for accounts that needed query coverage beyond their keyword list. ai max replaces this function, but with less advertiser control over landing pages. final url expansion sends users to whichever page google predicts will perform best. for lead-gen, this often means blog posts or about pages instead of contact forms. that is a conversion killer.

independent data from my own accounts and pmg data shows ~84% of testers see neutral or negative results when ai max is enabled without guardrails. the 16% who see positive results share common characteristics: robust negative keyword lists, clean conversion tracking, and url expansion disabled for lead-gen campaigns. one-click experiments are available to test ai max on a single campaign before committing. use them.

what this means for lead-gen buyers: dsa was the catch-all for accounts that needed query coverage beyond keywords. ai max replaces this function but with less advertiser control over landing pages. final url expansion sends users to whichever page google predicts will perform best. for lead-gen, this often means blog posts or about pages instead of contact forms. disable url expansion for lead-gen campaigns unless you have strict exclusion lists in place. prerequisites: smart bidding enabled, conversion tracking accurate, landing pages structured. one-click experiments available to test on a single campaign first.

for a complete deep-dive on ai max configuration, experiment setup, and lead-gen specific settings, see my ai max guide for 2026.

google ads api monthly release cadence: what changed in january 2026

google shifted the ads api from quarterly to monthly releases starting january 2026, announced september 4, 2025 on the google ads developer blog. the 2026 release calendar:

major versions sunset 12 months after launch. v23 sunsets in february 2027. google also launched a discord community server in july 2025 for advertising and measurement api discussions.

what this means for media buyers: new capabilities ship faster. programmatic control of ai-generated ad copy via text guidelines, ad-level scheduling, and unique_users metrics are already available. but you must pin your api version explicitly in ci pipelines or your reporting breaks silently. audit your integration this week.

performance max audience exclusions (march 2026)

google announced first-party audience exclusions for performance max at the end of march 2026 through the google business accelerate program. before this update, pmax audience signals were steering inputs, not hard exclusions. the algorithm could still show ads to people on your remarketing lists. now exclusions are rules.

you can exclude:

• remarketing lists
• customer match lists (email-based)
• site visitors (by duration or recency)
• app users (if you have app campaigns)

google also added budget projections (month-end spend forecast), demographic and network breakdowns in reporting, campaign-level negative keyword lists, and doubled search themes from 25 to 50 per asset group in the same release.

what this means for lead-gen buyers: stop wasting 15-25% of pmax budget on existing customers. typical improvement is 10-20% roas improvement in acquisition-focused campaigns. exclusion effect takes 24-48 hours to apply, so test on monday and measure by wednesday. customer match exclusions rely on email matching, so coverage is not perfect. but separating prospecting and retention into distinct campaigns is now a one-click operation instead of a budget leak.

for the overlap between pmax and ai max strategy, see my ai max deep-dive.

ask advisor: google’s cross-product ai agent (beta)

google announced ask advisor on may 20, 2026 at google marketing live. built with gemini capabilities, ask advisor works across google ads, google analytics, merchant center, and google marketing platform. it is the first unified google ai agent that operates across products, not just inside one surface.

current beta limitations:

• english-language accounts only
• not available in mcc accounts (per google ads help)
• beta quality output, not production-ready

ask advisor offers proactive recommendations, troubleshooting, campaign performance analysis, and text and image creative suggestions. it remembers business context across conversations, so you can ask follow-up questions without re-explaining your setup. thumbs up/down feedback is built in. but you are responsible for verifying suggestions before implementing them. google is clear on this point: ask advisor answers questions, you make decisions.

what this means for lead-gen buyers: first unified google ai agent that works across products, not just inside one surface. remembers business context across conversations. current limitations: beta quality, english only, no mcc support. not a replacement for strategy: it answers questions, you make decisions. worth testing for accounts spending $5k+/month where setup speed matters.

ai brief: natural language control for ai max

also announced may 20, 2026 at google marketing live, ai brief is powered by gemini and rolling out in beta for search campaigns. performance max and ai max for shopping support is coming soon. ai brief is the input layer. ai max is the optimization layer.

three guideline types:

• messaging guidelines: what ads should and should not say
• matching guidelines: queries to target or avoid
• audience guidelines: who to reach with what message

after input, ai brief shows sample assets and search term previews. chris monkman, google director for ads and ai experiences, called it “the answer to advertiser worries about control.” for lead-gen buyers, this means “never mention prices” or “avoid searches for in-person degrees” becomes enforceable guardrails instead of hoped-for behavior.

ai brief lets you steer ai max without learning gaql or api syntax. pair it with audience exclusions for the full control stack: ai brief steers the ai, audience exclusions hard-block existing customers, and weekly search term reviews catch what slips through.

what this means for lead-gen buyers: ai brief is the input layer. ai max is the optimization layer. “never mention prices” or “avoid searches for in-person degrees” becomes enforceable guardrails instead of hoped-for behavior. performance max and ai max for shopping support coming soon. pair with audience exclusions for full control stack.

content api sunset: august 18, 2026

the content api for shopping sunset on august 18, 2026. it is replaced by merchant api, a redesign with multiple sub-apis. progressive errors started september 1, 2026, meaning an increasing failure rate on content api requests. merchant api support in google ads scripts started april 22, 2026. v1beta of merchant api sunset february 28, 2026.

if you run shopping feeds via content api, the migration deadline has passed. after august 18, content api requests fail. agencies managing merchant feeds for clients should audit now. this update primarily affects shopping campaigns, but any lead-gen business running product feeds (auto parts, equipment, consumer goods) through the content api needs to verify migration status.

what this means for lead-gen buyers: if you run shopping feeds via content api, migrate to merchant api before august 18. after august 18, content api requests fail. after september 1, progressive errors (increasing failure rate). agencies managing merchant feeds for clients: audit now.

what to do this week (action items)

prioritize by urgency. some of these deadlines have already passed. others are months away.

priority 1 (this week)

• audit campaign types: identify which campaigns use dsa, aca, or campaign-level broad match. these will auto-upgrade in september. document current performance so you can measure the impact.
• verify content api status: if you run shopping feeds via content api, confirm migration to merchant api is complete. the sunset date was august 18.

priority 2 (this month)

• set up ai max experiment: enable ai max on one campaign with url expansion disabled. run for 4-6 weeks. measure qualified leads, not just conversions.
• add customer match exclusions: apply to your top 3 pmax campaigns. expect 10-20% roas improvement.
• pin api versions: update ci pipelines and reporting tools to pin explicit google ads api versions. prevent silent drift.

priority 3 (this quarter)

• test ask advisor beta: if eligible (english accounts, non-mcc), test for setup speed on new campaigns.
• build mcp read access: configure google ads mcp server for your ai agent stack (claude, gemini, cursor). read-only access to account data without manual exports.
• configure ai brief guidelines: write messaging, matching, and audience guidelines for brand-sensitive campaigns.

for campaign structure best practices that apply across all these updates, see my guides on google ads for auto detailers and the detailing lead generation playbook.

third-party performance data you should know

google publishes two headline numbers for ai max: 14% more conversions at similar cpa/roas (ai max activated vs not), and 7% more conversions when using the full feature suite vs search term matching alone. independent data paints a more complete picture.

lunio invalid traffic analysis (q2 2026):

• 114.5m google search clicks analyzed (49.5m ai max, 65m standard)
• ai max invalid traffic rose from 2.46% (q4 2025) to 5.28% (q2 2026)
• standard search invalid traffic: 3.72% → 2.98% → 3.07%
• ai max has 72% higher invalid traffic than standard search

smarter ecommerce (250+ retail campaigns):

• +13% conversion value, +16% cpa, 0% roas difference
• range: +42% to -35% (individual campaign variance is massive)

kickads portfolio:

• ai max roas was below exact-match in ~7/8 ecommerce accounts
• only won where exact match coverage was thin

additional 2026 updates:

• account-level placement exclusions: launched january 14, 2026, centralized exclusion list applies across pmax, demand gen, youtube, display
• language targeting removal: campaign-level language targeting for search and ai max removed late september 2026 (announced august 13, 2026)
• ai max for shopping beta: google claims +5% conversions/value at similar cpa/roas (retail, 2026)

these numbers do not mean ai max cannot perform. they mean the quality of your inputs determines the quality of your results. accounts with clean conversion tracking, robust negatives, and url expansion disabled for lead-gen see the best results. accounts without these foundations see waste. set expectations accordingly.

frequently asked questions