Between 1 and 30 September 2026, Google Ads will automatically upgrade Search campaigns that use campaign-level broad match or standalone automatically created assets into AI Max, migrating them in place with equivalent settings, according to Search Engine Land on 14 August 2026. The same report said Google stopped allowing creation of those legacy settings across the interface, Ads Editor and API on 3 August 2026, and that Dynamic Search Ads follow in February 2027.
For a UK owner running their own Google Ads, this is not a feature announcement. It is a change to what your money buys, arriving on a date you did not choose.
What exactly changes on 1 September 2026?
The detail that matters is which switches arrive on. PPC Land reported on 6 August 2026 that the two affected groups get different defaults. Campaigns migrating because of automatically created assets land with search term matching on and text customisation on, with final URL expansion off. Campaigns migrating because of campaign-level broad match land with search term matching on, text customisation off, and final URL expansion off.
That difference decides how much of your advertising a machine is now writing. Text customisation generates fresh headlines and descriptions. If your campaign is in the automatically created assets group, ad copy you did not write can appear against your business name from September onwards.
Why is Google forcing this now?
Google's own framing has been consistent since launch. The Google Ads blog post of 6 May 2025 described AI Max as three components: search term matching, asset optimisation including text customisation and final URL expansion, and a control layer covering locations of interest, brand controls, and search term reporting.
Read that structure carefully. Two of the three components loosen targeting. The third exists to tighten it again. The migration switches on part of the first by default and leaves the third for you to configure.
Which UK small businesses are affected, and which are not?
The businesses most exposed are the ones that accepted a Google Ads recommendation card at some point and never revisited it. Automatically created assets and campaign-level broad match were both routinely suggested as easy wins inside the interface, so a plumber, a dental practice or an independent retailer who clicked apply eighteen months ago may be carrying both without remembering.
If you have ever handed an account to a freelancer, inherited it from a previous agency, or built campaigns during a Google onboarding call, assume the legacy settings are present until you have looked.
What do the numbers actually say about AI Max performance?
PPC Land's 12 August 2026 report also cited a November 2025 study of retail campaigns finding conversions delivered at roughly 35 per cent lower return on ad spend against traditional match types, alongside earlier third party testing where the large majority of expanded impressions produced no conversions at all.
Those retail figures do not transfer cleanly to a UK service business with fifteen keywords and a £40 average enquiry value. What transfers is the shape of the risk: more reach, weaker average intent, and a widened gap between clicks and money.
What does this do to a £1,500 a month UK ads budget?
Take an illustrative Bristol plumbing business spending £1,500 a month at an average cost per click of £3.00. That buys about 500 clicks. At an 8 per cent enquiry rate, that is 40 enquiries. Closing 40 per cent gives 16 jobs at an average value of £220, so £3,520 of revenue and a cost per enquiry of £37.50.
Now assume search term matching shifts 30 per cent of those clicks to broader research queries that convert at 3 per cent rather than 8 per cent. The 350 core clicks still produce 28 enquiries. The 150 broader clicks produce about 4. Total enquiries fall to 32, cost per enquiry rises to roughly £47, and jobs drop to about 13. Revenue falls to £2,860 on identical spend.
The upside case is real too. If the broader queries convert at the same 8 per cent because they surface job types your keyword list never covered, enquiries rise above 40 and the feature pays for itself. Both outcomes are available. Which one you get depends almost entirely on whether you feed the system exclusions in week one.
Should you accept the upgrade or switch the legacy setting off?
If your Google Ads account records a conversion for every enquiry form and every phone call, then accept the migration and manage it. If it records nothing, or only counts page views, then turn the legacy setting off before the migration window and rebuild tracking first, because AI Max optimises towards the signal you give it and a blank signal produces expensive noise.
If you spend under about £500 a month, switch the legacy feature off and stay on tightly controlled exact and phrase keywords until February 2027. There is not enough volume for the system to learn anything useful, and one bad fortnight is a large share of the quarter.
If you spend £1,000 to £3,000 a month with a defined service area, accept the migration but set locations of interest and a negative keyword list before the window opens. This is the band where the expanded matching genuinely finds jobs, and also the band where two unattended weeks hurt.
If your average job value is over about £1,000, accept it and watch the search term report weekly. A single extra job covers a lot of wasted clicks. If your average job value is under £100, be far more conservative, because the maths leaves no room for a diluted query mix.
If you sell in one city but your website reads as national, fix the website before the migration. Expanded matching amplifies whatever ambiguity already exists in your pages, which is the same principle that governs how structured, well-scoped landing pages support paid media performance.
What breaks first when search term matching goes on?
An electrician who fits EV chargers starts matching queries about buying EV chargers. A salon starts matching training course searches. An accountant serving limited companies starts matching sole trader tax queries. None of those are errors in the ordinary sense. They are the system doing what it was told, which is to find adjacent demand.
The second failure is geographic. Location targeting and locations of interest do different jobs, and a campaign set to a radius can still serve against queries naming towns forty miles outside it. For any business where travel time decides profitability, that is money leaving on the first day.
The third is brand adjacency. PPC Land noted brand controls that let you decide whether your ads appear on queries containing brand names. If you have never touched that setting, it is worth opening before September rather than after.
How do you keep control after the upgrade?
Build the negative keyword list before the migration reaches you, not after. Start with the obvious commercial exclusions for your trade: free, DIY, jobs, courses, training, salary, wholesale, and the names of services you do not sell. Add every town outside your service area that shares a name with somewhere you do serve.
Keep pinned assets in mind. PPC Land reported that pinned responsive search ad assets remain respected for the broad match cohort because final URL expansion does not activate by default. If there are claims you must control for regulatory or trading standards reasons, pinning is the mechanism that protects them.
Check the search term report weekly rather than monthly for the first six weeks. Reporting now shows AI Max as a match type with source differentiation, which is what lets you separate what your keywords earned from what the expansion added. That separation is the whole basis of an honest decision in October.
What should you do this week, before 1 September?
First, list every Search campaign and record whether it uses campaign-level broad match, automatically created assets, both, or neither. Write the answer down. This is the only way to know which campaigns should change in September and which should not.
Second, export a fourteen day baseline now: spend, clicks, conversions, cost per conversion, and the top fifty search terms. Save it outside the platform. Without a pre-migration baseline, any October argument about performance is guesswork.
Third, verify conversion tracking actually fires. Submit your own enquiry form and place a test call. If neither registers, that is the priority, ahead of everything else on this list.
Fourth, build and apply the negative keyword list described above, and set locations of interest to match the area you genuinely serve.
Fifth, open brand controls and choose deliberately rather than accepting whatever default is present.
Owners who already document their processes will find this faster, which is the same operational habit that makes written process documentation pay off across a small business.
How will you know in October whether it worked?
Search Engine Land reported that the rollout is gradual across the month rather than a single switch, so your campaign might have changed on the third or the twenty seventh. Two campaigns in the same account can move on different days.
Judge on three questions. Did cost per qualified enquiry rise or fall against baseline? Did the search term report contain queries you would have paid for on purpose? Did the jobs actually booked change in the same direction as the conversion count in the interface? If conversions rose but jobs did not, the tracking is counting something that is not a customer, and that is a measurement problem rather than an AI Max problem.
If cost per qualified enquiry has risen by more than roughly a quarter and the search term report is full of off-target queries, tighten exclusions and re-measure rather than pausing. If it has risen and the queries look right, the problem is the landing page or the offer, not the matching.
What does this signal about the next twelve months?
The February 2027 Dynamic Search Ads migration is already scheduled, with reminders around 15 January 2027 and the ability to create new Dynamic Search Ad groups removed during that month, per Search Engine Land. Older API versions supporting legacy features reach sunset around September 2027. The direction is not ambiguous.
What still belongs to the owner is conversion data quality, exclusion lists, service area definition, landing page clarity, and the speed of follow-up after an enquiry lands. Those inputs decide outcomes now more than match type selection does. The same shift is visible across organic search, where the distinction between traditional SEO and AI-driven answer surfaces has changed which inputs a small business can genuinely control.
September 2026 is a small forced change with a useful side effect. It is a reason to open an account most owners have not looked inside for months, fix the tracking, write the exclusions, and set a baseline. Those tasks were worth doing before the migration was announced. They are simply harder to postpone now.
Google Ads Auto — FAQ
Do I have to move my Google Ads campaigns to AI Max?
Not directly, but the legacy settings are being retired underneath you. Search Engine Land reported on 14 August 2026 that campaigns using campaign-level broad match or standalone automatically created assets will be upgraded in place during September 2026, and that Google stopped allowing new versions of those settings to be created on 3 August 2026. You can avoid the automatic upgrade by switching the legacy feature off in the campaign before the migration reaches you, or by enabling AI Max deliberately on your own timetable so you control the settings rather than inheriting defaults. What you cannot do is keep the old configuration indefinitely. The practical choice is whether you arrive at AI Max on purpose with tracking and exclusions in place, or arrive at it by default in the middle of September.
Which of my campaigns will actually be affected?
Two groups. First, Search campaigns running automatically created assets, the setting that generates extra headlines and descriptions from your landing page. Second, Search campaigns using the campaign-level broad match setting. PPC Land reported on 6 August 2026 that the two groups get different defaults after migration. Automatically created asset campaigns arrive with search term matching and text customisation switched on and final URL expansion off. Campaign-level broad match campaigns arrive with search term matching on, text customisation off, and final URL expansion off. If your campaigns use exact and phrase keywords only, with no automatically created assets and no campaign-level broad match, nothing moves in September. Check each campaign individually rather than assuming, because settings often differ between campaigns built at different times.
Will AI Max make my ads more expensive?
It depends on your matching discipline, not on the feature itself. Google's own Ads blog claimed on 6 May 2025 that advertisers activating AI Max typically saw around 14 per cent more conversions or conversion value at a similar CPA or ROAS, rising to about 27 per cent for campaigns that relied heavily on exact and phrase keywords. PPC Land reported on 12 August 2026 that Google later revised the headline figure down to an average of roughly 7 per cent more conversions when the features run together. Independent numbers are less flattering, with PPC Land citing Lunio's finding that retail campaigns running AI Max faced substantially more invalid traffic. For a small UK budget the risk is not the CPC, it is a widened query set diluting your conversion rate.
What is search term matching and why does it matter most?
Search term matching is the part of AI Max that expands which queries your ads can appear on, using broad match plus keywordless technology to find searches your existing keywords miss. Google described it on 6 May 2025 as one of three components alongside asset optimisation and enhanced controls. It matters most because it is switched on by default for both migrating groups, according to PPC Land's 6 August 2026 report, and because it changes the composition of your traffic rather than just the cost. A plumber whose keywords previously matched emergency and repair phrases can suddenly match research phrases, DIY phrases, and job types that are not offered. Negative keywords and location controls are the levers that keep it useful.
How long should I give it before judging results?
Set a fixed review window rather than watching daily. A sensible pattern for a small UK budget is to record a fourteen day baseline before the migration reaches your account, then leave settings alone for the first fourteen days after it lands so the system has stable data, then compare. Judge on cost per qualified enquiry and jobs booked, not clicks or impressions. Search Engine Land noted on 14 August 2026 that migration rolls out gradually across September rather than on one date, so log the day your own campaigns actually change. If cost per qualified enquiry rises by more than roughly a quarter over the comparison window and the search term report shows off-target queries, act on exclusions rather than pausing the campaign outright.
What happens to Dynamic Search Ads?
They are on a later timetable. Search Engine Land reported on 14 August 2026 that Dynamic Search Ads migration begins in February 2027 and completes by the end of that month, with reminder notifications to affected advertisers around 15 January 2027 and the ability to create new Dynamic Search Ad groups permanently removed during February. The same article noted that older Google Ads API versions supporting the legacy features reach sunset around September 2027. If you run both broad match campaigns and Dynamic Search Ads, treat September 2026 as the rehearsal. Whatever you learn about negative keyword hygiene and location controls in the autumn is what you will need again in the new year, on a campaign type that has even less keyword-level restraint.
