AI in Google Ads: UK Statistics and Benchmarks (2026)

AI in Google Ads: the headline numbers for 2026

As of 2026, artificial intelligence is no longer a feature inside Google Ads, it is the operating layer. Google reports that Performance Max delivers an average 27% increase in conversions or conversion value at a similar CPA or ROAS, that Smart Bidding Exploration produced a 19% conversion increase in its own testing, and that AI Max for Search delivers an average 14% uplift. Independent research puts a hard number on the money involved: US AI-powered ad spend is forecast to reach $57bn in 2026, up 63% year on year, and UK paid search spend reached £17.9bn in 2025.

Those four numbers get quoted at each other constantly, usually without their sources attached. So the table below leads the page rather than closing it. Every row carries its publisher, its sample size where the publisher disclosed one, and the date the figure was published. Sort it, screenshot it, quote it.

Sources as listed in the table. Google figures are internal and unaudited; independent figures carry a disclosed sample size.

Two things are worth flagging before you use any of this. First, six of the ten rows come from Google, and Google has never submitted an uplift figure to independent audit or published a sample size alongside one. Second, none of the independent studies measures the UK specifically. Optmyzr, Smarter Ecommerce and WordStream all publish global or US samples. That gap is the reason this page exists.

How much of UK search spend now runs through AI-managed campaigns

UK paid search was worth £17.9bn in 2025, 44% of the country's £40.5bn digital advertising market, and grew 6% year on year (IAB UK Digital Adspend, published 2026). Nobody publishes the exact share of that £17.9bn that sits inside AI-managed campaign types, but the available proxies point the same way: Optmyzr found 51% of advertisers give Performance Max more than half their budget, and Smarter Ecommerce measured Performance Max peaking at 81.9% of ecommerce ad spend in May 2024.

The national picture

IAB UK's Digital Adspend study, published in 2026 and covering the 2025 calendar year, puts total UK digital adspend at £40.5bn, up 10%. Paid search takes £17.9bn of that, a 44% share, up 6%. Social reached £11.5bn, up 21%. Video reached £9.3bn, up 20%. Retail media reached £3.8bn, up 18%. IAB UK forecasts £44.7bn in 2026, a 10.3% increase, and £49.1bn in 2027.

UK digital adspend by channel, 2025. Paid search £17.9bn, social £11.5bn, video £9.3bn, retail media £3.8bn. Source: IAB UK Digital Adspend, published 2026 covering 2025. IAB UK's categories overlap, so they do not sum to the £40.5bn total.

For scale, UK digital advertising grew 10% in a year when UK GDP grew 1.4%. Whatever else automation has done to the mechanics of a campaign, it has not slowed the flow of money into the channel.

The forecast layer

IAB UK, working with Oliver Wyman and publishing on 10 March 2026, found that 48% of respondents named AI and automation as one of the most important forces shaping the sector over the next decade. The same piece references a forecast that one third of UK digital ad spend will be AI-driven by 2030.

What the platform accounts show

Alphabet's Q4 2025 earnings release, published 4 February 2026, reported Google Search and other revenue of $63.073bn for the quarter, up 17% year on year, within total Google advertising revenue of $82.284bn. EMEA revenue was $33.056bn, up 17%. Alphabet does not break out the UK, so this is a directional read rather than a national one.

Madison and Wall, reported via Business Insider on 2 April 2026, forecast US AI-powered ad spend at $57bn for 2026, up 63%, roughly 12% of a $475bn US ad market, with the other 88% of that market growing about 5%. The growth is attributed to Performance Max and Meta's Advantage+. On those numbers, AI-managed inventory is growing more than twelve times faster than everything else.

None of these figures measures "share of UK spend inside AI-managed campaigns". That number does not exist publicly. Anyone who quotes one is either estimating or reading a forecast as a fact.

Google's own published performance claims, and how to read them

Google publishes specific uplift figures for every AI product in Google Ads, and every one of them comes from Google's internal data with no independent audit. The current headline claims are 27% more conversions from Performance Max, 19% from Smart Bidding Exploration, 14% from AI Max for Search, and 15% from adopting AI Max alongside Performance Max. These are useful as a statement of what Google believes its own products do. They are not benchmarks.

Source: Google Ads Help and Google product announcements, as dated in the right-hand column. All figures are Google internal data with no disclosed sample size.

The advertiser results Google put on stage

At Google Marketing Live on 20 May 2026, Google named four advertisers and their results. These are selected case studies, not a sample, and should be read that way.

Source: Google, Google Marketing Live 2026, 20 May 2026.

None of that makes the figures useless. Google has more visibility into the auction than anyone else, and the direction of every figure is consistent. But a number without a sample size, a control group or an audit is a marketing claim, and it should be cited as one.

Smart Bidding: adoption and measured impact

Smart Bidding is now the default rather than a choice. Google's own most recent published figure, from Smart Bidding Exploration, is a 19% increase in conversions and an 18% increase in the number of unique search query categories that produce conversions, measured on global internal data between 11 March and 11 April 2025.

Google announced Smart Bidding Exploration on 21 May 2025 and described it as its biggest update to bidding in over a decade. The mechanism is that the bidder is allowed to spend against query categories it has not previously converted on, in exchange for a slightly looser adherence to the target. The 18% figure, the growth in unique converting query categories, is the more interesting of the two, because it says the gain came from reach into new demand rather than from squeezing existing demand harder.

Bidding does not sit on its own

Optmyzr's study of 24,702 campaigns, published 16 June 2025, found that bidding behaviour and budget allocation interact. Accounts with high Performance Max allocation recorded 652.03% ROAS, while the conversion rate and CPA advantages showed up in accounts that held Performance Max to 10 to 25% of budget. In other words the bid strategy that wins depends on how much of the account it is allowed to touch.

Smarter Ecommerce, analysing 4,000+ Performance Max campaigns across 500+ accounts through 2 April 2025, found roughly 95% of campaigns were using Maximise conversion value as of 2023, and that the median ROAS target rose from about 4.7 to about 6.0 over the analysis window. The split by strategy is the most useful cut they publish.

Maximise conversion value against Maximise conversions on median CPC, conversion rate and average order value. Maximise conversion value records €0.37 CPC, 2.22% conversion rate and €106.55 AOV; Maximise conversions records €0.25 CPC, 1.98% conversion rate and €22.84 AOV. Source: Smarter Ecommerce, n=4,000+ Performance Max campaigns across 500+ accounts, analysis through 2 April 2025.

The AOV gap is the finding people miss. Value-based bidding is not simply a better version of volume-based bidding, it selects for a different customer. A four to five times difference in average order value is not something a bid strategy creates in isolation, it reflects which advertisers choose which strategy and which baskets the algorithm learns to chase.

Automation has not killed the keyword

The State of PPC 2026 survey (n=1,306, fielded November to December 2025) found that 78% of professionals still use exact match keywords, and that exact match records the highest satisfaction of any targeting option at 77%. Automated bidding has not removed keyword control, it has changed what control means: you steer the input, not the bid.

Which brings the whole section to one point. An automated bid strategy is only as good as the conversion signal it is optimising against. Feed it a form-fill goal that fires on a thank-you page reload and it will confidently buy you duplicate leads at an excellent apparent CPA. This is why conversion tracking verification comes first in every what a proper account audit covers, and it is the first thing we check on a Google Ads audit. Signal quality is now the highest-leverage work in a paid search account, because it is the one input automation cannot repair on your behalf.

Performance Max: share of spend, ROAS and the control problem

Performance Max is used regularly by 50% of PPC professionals (State of PPC 2026, n=1,306), and 51% of advertisers give it more than half their budget (Optmyzr, n=24,702 campaigns). It reliably hits its ROAS target, delivering between 95% and 116% of target across 4,000+ campaigns measured by Smarter Ecommerce. The complaint is not performance, it is visibility: 56% of Performance Max users name black-box targeting as their main frustration.

Adoption and budget share

Optmyzr's 24,702-campaign study, published 16 June 2025, found 82% of advertisers ran Performance Max alongside other campaign types rather than on its own, and 51% allocated over half their budget to it. 71% used Search Themes, with results described as mixed or flat. 58% saw flat or slightly better performance with no exclusions at all. 57% supplied complete creative sets, and text-only assets often outperformed them.

That last pair of findings deserves more attention than it gets. Two of the most commonly recommended Performance Max optimisations, exhaustive exclusion lists and full creative sets, did not reliably improve results across a 24,702-campaign sample. That is not a reason to abandon either practice, but it is a reason to test them in your own account rather than treat them as settled.

Smarter Ecommerce's data, drawn from 4,000+ campaigns across 500+ accounts through 2 April 2025, adds the time dimension. Performance Max peaked at 81.9% of ecommerce ad spend in May 2024, then declined roughly 6% in total, losing about 0.65% of share per month into early 2025. The average number of Performance Max campaigns per account has nearly doubled since September 2022. Feed-based ads account for a median 90% of Performance Max cost, with a range of 74% to 97%. Dynamic Search Ads peaked at 21% cost share in Q2 2022 and have lost 5% share since Performance Max scaled.

Performance Max share of ecommerce ad spend. Only the two published data points are plotted: the reported peak of 81.9% in May 2024, and that peak less the reported total decline of roughly 6 percentage points into early 2025. Source: Smarter Ecommerce, n=4,000+ campaigns across 500+ accounts, analysis through 2 April 2025.

Performance against target

Smarter Ecommerce found Performance Max campaigns achieved between 95% and 116% of their ROAS target. Optimal performance appeared above 60 monthly conversions per campaign, with 30 monthly conversions as the minimum for reliable ROAS adherence. They found no evidence of learning transferring between campaigns, which undercuts the common argument for splitting a catalogue into many small Performance Max campaigns and expecting them to teach each other.

The control complaints

The State of PPC 2026 survey asked Performance Max users what frustrates them. The answers are almost entirely about visibility rather than results.

Top frustrations among Performance Max users: black-box targeting 56%, lack of granular control 48%, opaque attribution 41%, insufficient insights 33%, cannibalisation 32%, incrementality 31%, attribution modelling 27%. Source: State of PPC 2026, n=1,306 professionals across 50+ countries, fielded November to December 2025.

Despite that list, 43% plan to increase Performance Max budgets in 2026 and 38% still run Standard Shopping alongside it. The profession has settled on an uncomfortable equilibrium: the format works, nobody can see why, and the hedge is to keep a controllable campaign type running next to it.

AI Max for Search: what the 14% figure actually covers

When Google launched AI Max for Search campaigns on 6 May 2025, it published an average 14% increase in conversions or conversion value at a similar CPA or ROAS, rising to around 27% for campaigns that had been running mostly exact or phrase match keywords. Both figures come from Google's closed beta. Google's own Help documentation for AI Max carries no performance figures at all.

The 27% variant is the one worth thinking about. It is highest where the advertiser previously had the tightest control. That is simultaneously a statement about how much incremental matching AI Max unlocks, and a statement about how much control the advertiser hands over to get it. An account running mostly exact match has, by definition, the largest pool of unmatched adjacent queries sitting outside its keyword list. Opening that pool produces the biggest measured gain and the biggest loss of predictability at the same time.

The most-cited page on the topic publishes no data

Google's Help page "How AI Max for Search campaigns works" holds 207 referring domains, which makes it one of the most-linked documents about the product anywhere. It describes "incremental reach and performance" and "improved performance" in general terms and quantifies nothing. The single most authoritative citation target in this topic contains no statistics, which is a fair summary of the state of the evidence.

What the named results show

At Google Marketing Live on 20 May 2026 Google reported that adopting AI Max alongside Performance Max produced 15% more conversions at a similar ROAS, and named Lufthansa Group at a 24% return on ad spend increase and IKEA at a 65% lift in non-branded clicks with a 28% boost in incremental ROAS. The IKEA figure is the most informative of the set, because non-branded click lift is the mechanism the product is supposed to deliver, rather than a downstream outcome that could have several causes.

On the practitioner side, the State of PPC 2026 survey found 78% still use exact match keywords at 77% satisfaction, the highest of any targeting option (n=1,306, fielded November to December 2025). Keywordless expansion is being adopted as an addition to the keyword list, not a replacement for it.

Ads in AI Overviews and AI Mode

Google reports that AI Overviews now reach more than 2.5 billion monthly active users and that AI Mode has passed one billion monthly users, with AI Mode searches running on average three times as long as traditional searches (Google, Google Marketing Live, 20 May 2026). Ads are live in both surfaces, and Google states that 75% of shoppers say AI Mode helps them make faster, more confident purchasing decisions.

Longer queries are the mechanical argument for keywordless matching

Google also reported at Google Marketing Live 2026 that brainstorming queries are growing 30% faster than AI Mode searches overall. A query that is three times longer than a traditional search is a query a keyword list will not contain. That is the honest engineering case for AI Max and broad match: the shape of demand has changed faster than any manual keyword build can follow. It is also, inconveniently, an argument that only Google can verify.

The organic side of the same shift is documented in our first-party data on AI Overviews and organic traffic, which measures what happens to click-through when an AI Overview occupies the top of the page.

The documentation problem

Google's support page "About ads and AI Overviews" holds 302 referring domains, making it one of the most-cited pages on the topic. It is documentation, not data. Every one of those 302 links points at a page that describes ad eligibility and placement without publishing a single performance figure.

Context for why none of this reverses: Google reported at Google Marketing Live 2026 that Alphabet capital expenditure was $31bn in 2022 and is projected to reach roughly six times that in 2026. Infrastructure spending on that scale is not built to be switched off, and the ad products are the mechanism through which it is monetised.

What UK and European advertisers actually say

The largest recent survey of the profession, the State of PPC 2026 report, gathered 1,306 PPC professionals across more than 50 countries between November and December 2025, with UK and Ireland making up 12.8% of the sample and Continental Europe 45.8%. Its central finding is that 53% say paid media is harder than it was two years ago, up from 49% in 2024, and the most cited reason is platform opacity.

Is the job harder?

53% say harder, 16% say easier, and 31% say roughly the same. The reasons given, in order: less accurate conversion measurement at 53%, signal loss and privacy changes at 50%, increased competition at 43%, and less control over targeting and creatives at 43%. Black-box platform technology is the most cited single cause, and it is also the clearest example of secondary-reporting drift on this topic, which section eleven covers.

Why paid media got harder: less accurate measurement 53%, signal loss and privacy 50%, more competition 43%, less targeting control 43%. Source: State of PPC 2026, n=1,306, fielded November to December 2025.

AI in the daily workflow

66% of respondents treat large language models as their primary problem-solving resource. 59% use ChatGPT for ad copy, 39% for keyword research and 34% for script writing, saving around five hours a week on average. 22% report vibe coding. Agent adoption sits at 21% overall, rising to 28% among accounts spending over $3m a month. The blockers are quality and accuracy concerns at 58% and reliability and trust at 43%.

Primary AI assistant among PPC professionals: ChatGPT 56%, Gemini 17%, Claude 13%. Source: State of PPC 2026, n=1,306, fielded November to December 2025.

43% name AI and automation as a 2026 priority, down from its top-ranked position in 2024. Read that as a maturity signal rather than a retreat: the technology has stopped being a project and started being furniture.

Priorities and operating reality for 2026

Source: State of PPC 2026, n=1,306 professionals across 50+ countries, fielded November to December 2025. Average of 5.4 scripts per account reported.

The UK SMB view

Amazon Ads research covering 4,100 respondents across 14 countries, published 7 August 2026, found 68% of SMBs said AI advertising supported growth and 75% see AI as crucial to future competitiveness. 88% plan to use AI to create adverts, 52% said AI supported expansion into other countries, and 73% of UK SMBs said AI advertising tools make streaming TV affordable and accessible. Small advertisers report the benefit as access, not efficiency, which is a different claim from the one the platforms make.

Cost and conversion benchmarks in an AI-managed account

The most widely cited Google Ads benchmark set for 2026 comes from WordStream, drawn from 13,474 search campaigns across 23 industries between 1 April 2025 and 31 March 2026, published 19 May 2026. It reports a median click-through rate of 6.64%, a median cost per click of $5.42, a median conversion rate of 8.18% and a median cost per lead of $66.69. Those campaigns are all US-based. UK costs differ.

Source: WordStream Google Ads Benchmarks, 13,474 US search campaigns across 23 industries, 1 April 2025 to 31 March 2026, published 19 May 2026.

Conversion rate rose from 2.70% to 8.18% while cost per lead rose from $59.18 to $66.69 across the same decade. Source: WordStream Google Ads Benchmarks, n=13,474 US search campaigns, published 19 May 2026.

The 2026 movement

WordStream reports that conversion rate increased for 87% of industries year on year, and that cost per lead fell for the first time in five years. The largest conversion rate gains were in beauty and personal care, up 32.34%, and personal services, up 26.69%. The largest cost per click falls were in education and instruction, down 22.79%, and beauty and personal care, down 18.95%. The largest cost per click rise was in real estate, up 27.27%.

Two defensible readings of the same decade

Over a decade in which automated bidding went from optional to default, conversion rate tripled while cost per lead rose only 13%. That is the strongest available quantitative case that automation improved match quality: the platform got substantially better at putting the ad in front of someone who would act, and the cost of acquiring an action barely moved in nominal terms.

It is also the weakest case, because the same decade included a wholesale change in conversion tracking, attribution modelling and what the industry counts as a conversion. Modelled conversions, enhanced conversions and data-driven attribution all attribute more actions to paid search than the 2016 methodology did. Some of the tripling is better matching and some of it is better counting, and no published dataset separates the two. Both readings are defensible and this page will not pick one.

Our own UK medians from the accounts we manage are set out in the next section.

First-party benchmarks from the UK accounts we manage

Across the 34 live UK Google Ads accounts Visionary manages, 58% of media spend now runs through Performance Max, and median achieved ROAS came in at 101% of target. Those two numbers are the short answer to most of the questions raised by the third-party data above, and neither of them has been published anywhere else for a UK agency book.

Portfolio composition

Across the portfolio, 58% of managed spend now runs through Performance Max. Account level share ranges from 9% to 91%, which is a wider spread than any published benchmark suggests. The portfolio figure sits well below Smarter Ecommerce's 81.9% peak for ecommerce ad spend, and that gap is mostly composition: our book carries lead generation accounts that no ecommerce-only sample contains.

87% of active campaigns in the portfolio run on an automated bid strategy. Target ROAS accounts for 34%, Maximise conversion value 21%, target CPA 19% and Maximise conversions 13%. Manual and enhanced CPC now account for 13%, and those are concentrated in low volume brand campaigns and in local service campaigns where monthly conversion counts are too thin for an algorithm to learn from.

Performance Max at scale, inside one account

The largest ecommerce account in the portfolio runs 63 separate Performance Max campaigns split across 22 brand groups, on a monthly media budget in the £40,000 to £60,000 band, and has been under management for 26 months. That structure is unusual. Most advice on Performance Max assumes a handful of campaigns and a single asset strategy, and almost none of it addresses what happens when a catalogue is deep enough that 60 campaigns is the only way to hold margin apart by brand.

Performance against target

Across 148 Performance Max campaigns, median achieved ROAS came in at 101% of target. Smarter Ecommerce reported a 95% to 116% range across more than 4,000 campaigns, so our portfolio sits inside that band. The more useful reading is that Performance Max is reliable at hitting whatever target it is given, which makes the target itself the thing worth arguing about.

The conversion volume threshold, tested

Splitting our Performance Max campaigns by monthly conversion volume, 26% sit above 60 conversions, 33% between 30 and 60, and 41% below 30. Median ROAS against target was 114%, 102% and 89% respectively. Smarter Ecommerce's 30 and 60 monthly conversion thresholds hold up in our data, and the penalty below 30 is larger than we expected.

Source: Visionary Marketing, n=148 Performance Max campaigns across 34 UK accounts, trailing 12 months to August 2026.

The budget concentration test

Where Performance Max holds under 25% of an account's budget, median conversion rate was 4.1% at a CPA of £38. Where it holds over half the budget, median conversion rate was 3.2% at a CPA of £47. That is the same direction Optmyzr found across 24,702 campaigns, and it is a real effect in our book rather than a sampling artefact. It is also not an argument for starving Performance Max. 35% of our accounts run it above half of budget, against Optmyzr's 51%, and several of those are the accounts with the deepest catalogues and the least useful keyword data.

Source: Visionary Marketing, n=34 UK Google Ads accounts, trailing 12 months to August 2026. The remaining 36% of accounts sit between the two bands.

AI Max and keywordless expansion, kept or rolled back

UK cost per click by sector

Median CPC across our UK portfolio was £1.84 over the trailing 12 months. By sector, legal services ran at £4.62, B2B software at £3.95, home improvement at £2.71 and ecommerce retail at £0.68. WordStream's 2026 benchmark of $5.42 is drawn from 13,474 US campaigns and does not translate directly, and converting it into pounds produces a number that is arithmetically correct and commercially useless for a UK advertiser.

The audit finding

Of the 29 Google Ads accounts we audited in the last 12 months, 72% had duplicate or overlapping conversion actions counted as primary. Every one of those accounts was feeding a Smart Bidding algorithm on an inflated conversion count, which means the algorithm was optimising confidently towards a number that did not exist. This is the least glamorous finding on this page and the most consequential one. Automated bidding is only as good as the signal underneath it, and the signal is broken in roughly three accounts out of four.

The mechanics of finding it are covered in our Google Ads audit process.

Where the data is contested

Four of the most quoted claims about AI in Google Ads are disputed, inconsistent, or measure something narrower than they appear to. Google has published its Performance Max uplift as 18%, 27% and 15% in different places. Optmyzr's data and Google's guidance point in opposite directions on how much budget Performance Max should hold. And the two largest independent studies of Performance Max adoption disagree about whether it is growing or shrinking.

The common thread is sample composition. Almost every disagreement on this page dissolves once you ask which advertisers were measured, over what period, and against what comparison. Very little of the contested data is wrong. Most of it is being applied to populations it never described.

What this means for UK advertisers in 2026

The evidence supports three practical conclusions. Automated bidding is no longer a decision, it is the baseline, and the useful work has moved to the quality of the conversion signal feeding it. Performance Max reliably hits its ROAS target but does not tell you where the money went, so incrementality testing has become the highest-value skill in the account. And the benchmark numbers most advertisers quote at each other come from US campaign samples that do not describe a UK account.

One last point about where the discipline is heading. AI Mode queries run three times longer than traditional searches and 2.5 billion people now see AI Overviews monthly, which means the same retrieval-and-summarisation layer that reshaped paid search is reshaping organic discovery too. If you are planning for one, plan for both: our work on generative engine optimisation covers the organic half of that shift, and our Google Ads management pages cover how we apply the paid half in live accounts.

Frequently asked questions