H2 2026 Google Ads Playbook: What We’re Changing Now the AI Features Are Real

Last year’s playbook was about surviving CPC inflation. This one is about something different: between Google Marketing Live in May and the wave of releases that followed, Google shipped more AI features in six months than in the previous three years. Some of them are genuinely useful. Some of them are Google asking for more control of your budget. This is what we are adopting, what we are testing carefully, and what we are leaving alone in H2 2026, across the accounts we manage.

The context (and why it matters)

Three things make this half-year different.

First, the measurement stack got rebuilt. Since June, Enhanced Conversions is one unified module for web and leads, Data Manager connects your email and CRM platforms directly, and Tag Gateway serves Google tags from your own domain (Google cites up to 14% more measured conversions from that alone). The pattern is clear: every new bidding feature this year assumes your first-party data is flowing into the account.

Second, AI Max stopped being optional for some accounts. It left beta in April. Google has pushed the forced Dynamic Search Ads migration back to February 2027, but automatically created assets and campaign-level broad match still flip to AI Max in September 2026, and new DSA creation ends in January 2027. If you run DSAs, this half is your window to migrate on your own terms rather than Google’s.

Third, ads are moving inside AI answers. Google is testing three ad formats in AI Mode, which changes what a good ad asset even looks like.

What we’re changing (and how you can copy it)

1) Measurement first, before any AI feature gets switched on

• Unified Enhanced Conversions on every account. This is the single biggest quick win in tracking right now: Google’s own figures put the median uplift at around 5% more conversions on Search and 17% on YouTube. We check diagnostics 48 hours after the first import and want match rates above 50%; a well-implemented UK setup typically lands between 55 and 70%. Since April, Google accepts tag, Data Manager, and API signals together and deduplicates them, so on accounts where measurement carries real money we run the Google Tag and the Data Manager API import side by side for redundancy. Full setup in our unified setup guide.
• At least one source connected in Data Manager (Mailchimp, Klaviyo, ActiveCampaign, or your CRM). Google cites an average 26% incremental ROAS for advertisers connecting offline and app data through it; we treat that as directional, but the direction is right.
• The offline conversion loop for lead gen: capture hashed email and phone at form submission through Enhanced Conversions for leads, with the gclid alongside, persist it in the CRM, and import quality outcomes back as leads progress. Plain gclid file imports are now legacy, and API uploads moved to the Data Manager API in June. On the accounts where we run this loop properly it routinely cuts wasted spend by 30 to 40%. Google’s new built-in Leads CRM will do some of this automatically, but the rollout is staged through 2026 and it only covers Google-hosted form leads, so we build the loop ourselves.

2) AI Max: adopt by experiment, never account-wide

• One campaign or a proper experiment, 30 days, controlled budget. Not a whole-account flip. Brand exclusions and negative lists go in before launch, then we pull the search terms report weekly for the first month and negate immediately as junk appears. The new search term reporting has known gaps, so weekly review is the floor, not the ceiling.
• Write the AI Brief before launch, including the phrases you never want in your ads (we list at least three to five). The AI writes creative either way; the brief is your only steering wheel. AI Brief is still rolling out, so if your account does not have it yet, the older text guidelines control does the same job for banned phrases.
• Expectations set by independent data, not the headline. Google’s 7% uplift claim is real but narrowly qualified, and practitioner studies put the genuine account-level incremental gain nearer 3%, often with higher CPAs in the first weeks. That is exactly why we experiment rather than believe.
• DSA accounts migrate now, on our timetable rather than February 2027’s, and before the September flip of automatically created assets and campaign-level broad match.
• Campaigns with fewer than about 30 conversions a month wait (the feature only becomes reliable above 100), and budget-capped campaigns do not enable it at all: there is not enough data for the machine to learn from, and the broader matching burns budget faster than it teaches.
• Retail gets the same treatment with AI Max for Shopping, which is still in beta: one lower-priority campaign or single category first, and only with a clean feed and strong conversion tracking underneath it.

3) Budgets: new tools, old discipline

• Campaign total budgets for anything time-bound: launches, sales, seasonal pushes. Never for always-on campaigns, because you cannot switch budget types mid-flight. We check spend at the 25% and 50% marks of every flight.
• Demand-led pacing gets a few weeks to prove itself before we intervene. Then we review the spend distribution report against impression share lost to budget by day of week.
• The missed opportunity report is a conversation starter, not a spend justification. The report runs on Google’s value model, not your margins. Our house rule: scale only when lost impression share (budget) sits consistently above 20 to 25%, CPA is at or below target, and conversion rates are stable. Scaling on a broken CPA just scales a loss.

4) Full-funnel bidding, gated on CRM readiness

• Journey-aware bidding only enters an account that passes three gates: gclid capture persisting in the CRM, multiple funnel-stage conversion actions imported (we want at least three), and the bid strategy seeing 30 to 50 conversions a month with roughly three months of clean CRM history behind it. This is not a switch you flip and walk away from.
• AI lead intent scores only where the CRM discipline exists to verify them against real sales outcomes.
• If the CRM is a mess, we fix that first. That is the honest sequence, and it is why we build CRM pipelines alongside campaigns.

5) Stop paying to reacquire your own customers

• Where existing customers eat a meaningful share of paid conversions (our rule of thumb: a quarter or more), we are testing the new prospects mode alongside standard campaigns. It is stricter than the older new-customers-only setting: it also excludes brand searchers, site visitors, and past ad engagers, and it spans PMax, Search, Shopping, and Demand Gen. It needs a Customer Match list of at least 1,000 active customers (fresh export with email, phone, and postal address, refreshed monthly) and the new_customer tag parameter enabled.

6) Demand Gen: pick the updates that fit your business

• Google Maps Promoted Pins now, for anyone with a physical location or local service area.
• Checkout links on YouTube Shorts for e-commerce in supported markets; Google reports a 6% conversion uplift.
• AI-assisted setup pre-filled from PMax is a starting point, not a final answer. Review every pre-filled choice.
• View-through conversion bidding only after a conversion audit, otherwise you are optimising toward noise. It is YouTube inventory only for now, and note that Demand Gen with it enabled on Discover placements moves to CPM billing from mid-July. The full rundown is in our Demand Gen 2026 guide.

7) Creative that survives AI Mode

• Every headline and description has to work standalone, because AI Mode lifts assets out of your carefully ordered ad and drops them into a conversation.
• Offers get specific. “Three months for the price of two before 30 June” beats “great value” in an AI-quoted answer.
• Structured data on the site for services, reviews, and pricing, so the AI formats have something authoritative to pull from. This is where AEO and paid search stop being separate disciplines.

Your H2 routine (copy-paste into your runbook)

Weekly

• Search terms review on every AI Max experiment; negate the top wasters
• Budget re-allocation by marginal CPA/ROAS, 10 to 15% to winners
• Two new headlines tested, written to stand alone in AI Mode

Monthly

• Import CRM outcomes; check Enhanced Conversions match rate is still above 30%
• Refresh Customer Match lists (prospects mode depends on it)
• Review each AI feature experiment against its 30-day verdict date: adopt, extend, or kill

Quarterly

• Incrementality check on anything Google claims is working (geo split or holdout)
• Feed and structured data audit (retail: before the AI Max for Shopping rollout widens)
• Funnel review: are the journey-aware bidding gates passed yet?

What we’re not doing (and neither should you)

• Flipping AI Max on account-wide because Google’s interface suggests it
• Scaling budgets off the missed opportunity report while CPA is above target
• Running campaign total budgets on always-on activity
• Enabling journey-aware bidding on top of a messy CRM
• Taking Google’s uplift claims (+7% here, +26% there) on faith instead of verifying against our own conversion data

QA checklist (before any AI feature goes live)

• Tracking: Enhanced Conversions live, match rate above 30%, no duplicate conversion actions
• Guard rails: AI Brief written, banned phrases listed, brand exclusions set where relevant
• Experiment design: one campaign, controlled budget, a calendar reminder for the 30-day verdict
• Exit plan: know what “kill it” looks like before you press enable

Why this works now

Google’s AI features genuinely perform when they are fed clean data and held to experiment discipline, and they genuinely burn money when they are switched on blind. The accounts winning in H2 2026 are not the ones adopting everything or refusing everything. They are the ones sequencing it: measurement first, experiments second, rollout only on evidence. That sequence is the whole playbook.

If you want this level of discipline applied to your own account, that is what our Google Ads management is built on, or start with a Google Ads audit and we will show you exactly where your account stands against this playbook.