Loading…
Loading…
Google is replacing Dynamic Search Ads with AI Max for every account by September 2026. UK brands that prepare now keep control of budget and brand. The ones that wait get whatever the algorithm decides.
On 15 April 2026, Google confirmed what the industry had suspected since late 2025: Dynamic Search Ads are being retired. Starting in September 2026, every DSA campaign across every Google Ads account will be automatically upgraded to AI Max for Search — whether you opt in or not.
This is not a minor rebrand. AI Max changes how queries are matched, how headlines are written, which URLs get used, and how much the system leans on first-party signals. If your account currently depends on DSA for long-tail coverage, that coverage will look and behave differently in a few months. The settings you relied on to keep things predictable — domain pinning, page feeds as the only source of truth, manual URL exclusions — either move, change name, or disappear.
We are publishing this guide because the migration window is short, the documentation from Google is thin in places, and most UK advertisers we speak to have not yet opened AI Max in their accounts. The brands that run a controlled test in April and May will land September in good shape. The brands that wait will inherit whatever defaults Google chooses for them.
AI Max for Search is a campaign setting, not a new campaign type. It sits inside standard Search campaigns and bundles three features that Google has been building separately for the last eighteen months: search term matching (a broader match system that behaves more like a signal than a rule), text customisation (dynamic headlines and descriptions generated from your landing pages and asset library), and final URL expansion (the system can send traffic to any relevant URL on your domain, not just the ones you nominated).
When you switch AI Max on, your Search campaign stops behaving like a traditional keyword-plus-ad-copy structure and starts behaving like a topic-plus-creative-pool structure. You still provide keywords, but Google treats them as one input among many — alongside your landing pages, asset library, customer match lists, and offline conversion data. The engine decides what to match, what to show, and where to send the click.
The closest mental model is Performance Max for Search only: you give the system intent signals and creative raw material, and it makes the small decisions for you. The critical difference is that AI Max keeps the reporting infrastructure Search advertisers expect — you can still see search terms, landing pages, device and location performance. Performance Max hides all of that. For UK advertisers burned by PMax opacity, this is the single biggest reason to take AI Max seriously.
DSA worked by crawling your website or a page feed, matching queries against the content it found, and writing a headline on the fly. It was blunt but predictable. AI Max keeps the crawling behaviour but adds three layers on top, and each one changes how the campaign spends money.
First, query matching is broader. DSA matched queries to pages it had crawled. AI Max matches queries to inferred intent, which means it can trigger on terms your pages do not explicitly mention — brand adjacencies, problem-framing queries, comparison terms. In our beta accounts, this expanded coverage by 18–35% in terms of unique search queries, but around a quarter of the new queries were commercially irrelevant without proper brand and negative controls.
Second, text customisation writes more of the ad than DSA ever did. DSA auto-generated headlines. AI Max auto-generates headlines and descriptions, can swap in customer review snippets, can use your call-to-action library, and will test multiple variants automatically. The upside is relevance. The downside is brand drift — without text guidelines configured, the system will confidently write headlines your brand team would never approve.
Third, final URL expansion is on by default. This is the feature that most surprises UK advertisers when they open AI Max. Even if you set a final URL at the ad group level, the system may send the click to a different page on your domain if it predicts a better conversion outcome. This is powerful for large catalogues and documentation-heavy sites. For brands with a tightly curated funnel, it is a control you must explicitly lock down.
AI Max ships with more advanced controls than DSA ever had. These are not optional for UK brands — they are the difference between a campaign that respects your brand and budget guardrails and one that quietly drifts.
Brand controls let you specify brands to include and exclude at the campaign level. If you sell third-party products, whitelist the brands you are authorised to advertise. If you run a pure-play brand, exclude competitor names from ever appearing in generated headlines or search term matching. We have seen AI Max confidently generate competitor-name headlines in beta accounts that had no brand control configured.
Location controls let you restrict AI Max to specific geographies with finer granularity than standard location targeting. For UK-only advertisers, this matters because AI Max's intent matching can pick up queries that are technically UK-based but commercially irrelevant (for example, a UK IP researching a US-only product). Tighten the location signal to regions you actually service.
Text guidelines are effectively a brand voice prompt for the text generator. You can specify tone, required phrases, forbidden phrases, and regulatory disclaimers. UK advertisers in regulated sectors (financial services, legal, health) should treat this as mandatory — it is the only place to tell the system not to generate claims that would breach FCA, SRA, or MHRA guidance.
URL inclusions and exclusions override final URL expansion. If there are sections of your site that should never receive paid traffic (careers pages, legal pages, thin category pages), add them as exclusions before you launch. If you want traffic locked to a specific funnel, add the inclusions and remove everything else.
Google's AI Max documentation is written for a global advertiser. Several of the assumptions in that documentation break in the UK market, and the Google Ads UK team has been inconsistent about flagging them.
Currency and pricing signals drift. AI Max pulls price signals from your pages and uses them in generated ad copy. For UK advertisers running multi-currency sites (especially e-commerce brands with GBP, EUR and USD storefronts), we have seen AI Max pull USD prices into UK-facing ads because the schema on a default-language page returned a USD value. Audit your Product schema and OG price tags before enabling AI Max, or lock text guidelines to require '£' in pricing references.
Regional dialect and spelling matters for CTR. AI Max's text generator defaults to American English spellings unless your landing pages use British English consistently or text guidelines explicitly require it. 'Optimize' versus 'optimise' is a small thing, but across thousands of impressions it shows up as a measurable CTR gap — UK searchers trust ads that match their spelling conventions.
UK consumer protection rules are stricter than Google's global review. The CMA and ASA expect specific claim substantiation for comparative statements, 'best' or 'leading' adjectives, and discount framing. AI Max will generate these phrases by default if your page contains them. Add forbidden-phrase text guidelines for superlatives you cannot substantiate, or expect ASA complaints to show up from AI-written headlines you never approved.
GA4 and server-side tracking readiness affects AI Max more than DSA. The system's learning phase relies heavily on conversion signal quality, and UK accounts running basic client-side pixel tracking will see slower optimisation and worse results than accounts with enhanced conversions, Consent Mode v2 properly configured, and offline conversion imports. If your tracking is behind, fix it before migrating — AI Max amplifies both good and bad signal quality.
Google will auto-upgrade DSA campaigns in September 2026. That is the forced migration. What we are doing for every UK client between now and August is running a controlled voluntary migration in three phases, so we own the learnings before Google owns the switch.
Phase one (April–May): open AI Max in a single, low-stakes campaign in each account. We pick a campaign that is not business-critical — often a secondary product line or a market-testing campaign — and we enable AI Max with all the controls above configured from day one. We run it in parallel to the equivalent DSA or standard Search campaign, with matched budgets, for three to four weeks. The goal is not to prove AI Max wins; the goal is to see how it behaves in our specific account with our specific brand rules.
Phase two (June–July): migrate the next tier of campaigns once we have calibrated the brand, location and text guideline settings in phase one. This includes DSA campaigns that are performing adequately but not exceptionally — the ones where we are willing to accept 10–15% volatility during the learning phase. We keep high-performing core Search campaigns on the old structure until phase three.
Phase three (August): migrate core revenue campaigns last, with a two-week rollback plan in place. By August we have 12 weeks of AI Max data in the account, refined controls, and a clear picture of which conversion signals are driving the model. Core campaigns migrate with confidence, not with hope.
The principle behind the phased approach is boring but important: we want to be calibrating AI Max with non-critical budget in April, not calibrating it with our largest revenue driver in October when something goes wrong.
Once AI Max is live, the question becomes: what role does it play alongside your other campaigns? The advertisers we see struggling are the ones who expect AI Max to do everything. The ones who succeed use it as one layer in a hybrid structure.
The structure we recommend: core brand and top exact-match commercial terms stay in traditional Search campaigns with manual match types and tight ad copy. AI Max handles the long-tail and discovery layer — the queries you could never afford to build keyword-by-keyword, and the intent signals that traditional keyword targeting misses. Performance Max stays for catalogue and asset-driven campaigns where Shopping and Display placements add value.
This mirrors the approach we wrote up in our £42K Google Ads experiment — automation at the edges, manual control at the core. AI Max fits that framework cleanly: it replaces DSA as the long-tail layer, without replacing the manual Search campaigns that protect brand terms and high-intent commercial queries.
The trap to avoid is letting AI Max and Performance Max compete for the same queries. We use campaign-level negative keyword lists to keep AI Max off the terms Performance Max is already capturing, and vice versa. Without this separation, you end up paying Google to auction against yourself, which is exactly what the platform would like you to do.
AI Max will not look stable for at least three weeks. The learning phase is noisier than Target CPA's learning phase and it manifests differently — you will see CPCs spike, search term relevance wander, and conversion volume move up and down day-to-day. This is normal. What matters is the trend across weeks, not the noise within weeks.
The four metrics we track obsessively during the first 30 days: search term quality (what percentage of triggered queries are commercially relevant — we score this manually on a weekly sample of 100 queries), URL expansion accuracy (what percentage of clicks are landing on the pages we actually want them on), generated headline quality (weekly review of top 20 generated headlines by impression volume, scored against brand guidelines), and incremental conversion lift (comparing AI Max's reported conversions to an incrementality-adjusted number using holdout tests or pre-post analysis).
If search term quality is below 70% after three weeks, the issue is usually brand and negative controls. If URL expansion accuracy is below 80%, you need tighter URL inclusions. If generated headline quality is low, your text guidelines are not specific enough. If incremental lift is below the reported conversion lift, the system is harvesting demand you already owned — same PMax cannibalisation dynamic, different campaign type.
We have heard this from several UK advertisers: if Google is going to migrate everything in September anyway, why spend time and budget testing AI Max now? It is a fair question with a clear answer.
The forced September migration will inherit your current DSA settings and Google's recommended AI Max defaults. The defaults include final URL expansion on, no brand controls configured, no text guidelines set, and no URL inclusions or exclusions. In other words, the forced migration will turn your carefully controlled DSA campaigns into wide-open AI Max campaigns overnight.
Advertisers who voluntarily migrate now have the opportunity to configure every control before the system starts spending at scale. Advertisers who wait will wake up in September to campaigns that look nothing like what they approved, and they will spend the first two weeks of Q4 reactively adding controls while the system learns badly.
For UK e-commerce and B2B brands with material Google Ads spend, the cost of a controlled April–August migration is a fraction of the cost of a reactive September migration. That is the trade we are making with our clients, and the one we would recommend to any UK advertiser running more than £5,000 per month through DSA campaigns.
We are running phased AI Max migrations for UK clients between now and September 2026. Book a free audit and we will show you exactly which campaigns to migrate first, which controls to lock down, and how to avoid the default settings that cost most advertisers their first Q4.