AI Advertising · Pillar Guide

AI and Advertising in 2026: What Changes for Brands, Agencies and Platforms

A dated, platform-by-platform field guide to what Meta, Google and ChatGPT now decide for you, what you still own, and how to check that the results are real profit.

Ajay Kumar • •16 min read
Illustration of automated ad controls run by robotic arms while a founder keeps hold of the offer, creative and data levers

Quick answer: what does AI change in advertising?

AI in advertising now means the ad platforms make most delivery decisions for you. Meta, Google and ChatGPT choose audiences, placements, budget splits and creative combinations by default. Your job has moved to the inputs: the offer, creative variety, clean first-party conversion data and exclusions. You also need to test whether the results are real profit.

AI and advertising in late 2026: the platforms now decide by default

AI in advertising is the use of machine learning and generative AI to choose who sees an ad, where, at what bid and with which creative. In 2025 and 2026 this stopped being optional. It is now the default on every major platform.

Three shifts that affect your budget

For a brand spending ₹10 lakh to ₹1 crore a month, the real question is not whether to use AI. It is which controls still matter, where the money leaks, and how to check the results are real.

This pillar guide covers brands, agencies and the platforms themselves. Separate guides go deeper on agencies, creativity, ethics and data.

Timeline: the AI ad changes that move spend, 2024-2026

Most guides describe AI in general terms. The table below lists the dated changes that actually move money, taken from platform documentation and official announcements.

DatePlatformChangeWhat it means for you
3 Oct 2024GoogleAds in AI Overviews launch for US mobile usersSearch budgets start serving inside AI answers
2025MetaAdvantage+ Shopping renamed Advantage+ SalesGuides using the old name often use old rules
2025 to Feb 2026MetaExisting-customer budget cap removed (API field deprecated with v25.0)You can no longer cap spend on past buyers by percentage
1 Apr 2025India tax6% equalisation levy on online ads abolishedOne less cost on foreign ad platform bills
May 2025GoogleAI Max for Search announcedAn AI layer added on top of existing Search campaigns
16 Dec 2025MetaMeta AI chats used to personalise ads (not UK, EU, South Korea)What people ask Meta AI shapes the ads they see
9 Feb 2026OpenAIChatGPT ads pilot for logged-in adult Free and Go users in the USA new paid channel opens
5 May 2026OpenAISelf-serve Ads Manager, CPC bidding, minimum spend removedAny brand can test without a large commitment
Aug-Sept 2026OpenAIUK, India, Middle East, Europe, South-East Asia; 60+ countries by 23 SeptAvailable in most markets you sell in
Sept 2026GoogleText-customisation campaigns auto-upgraded to AI Max; campaign-level broad match campaigns too, per Search Engine LandCheck Search campaigns for settings you did not choose
5 Oct 2026OpenAIVisual ads during image generation (US test), DoubleVerify and IAS pilotsIndependent brand-safety checks reach the channel

One more change matters if you use a tool or agency that builds campaigns through the API. Old-style shopping campaigns can no longer be created, duplicated or updated through the Meta Marketing API. The block started with v25.0 and has applied to all API versions since 19 May 2026. Anyone still building them needs to move.

Meta: what Advantage+ decides and what you still control

On Meta, the algorithm now decides audience, placements and how budget splits across ad sets. You still own the offer, the creative inputs, the conversion data, exclusions and basic age or gender limits.

Advantage+ is now a state, not a campaign type. A sales, app or leads campaign shows Advantage+ as on only when campaign budget, audience and placements are all on. Narrow one too far and that section switches to Advantage+ off, often without anyone noticing.

Advantage+ Sales also allows multiple ad sets, each reportedly with up to 50 ads (in line with Meta's general ad set limit), plus basic exclusions such as age or gender. Our guide to optimising Advantage+ Sales campaigns covers the setup in detail.

DecisionWho decides nowWhat you can still do
AudienceMetaExclusions, age or gender limits, creative that signals who it is for
PlacementsMetaExclude only with clear data, knowing it switches Advantage+ off
Budget splitMeta (campaign budget)Ad-set structure, separate campaigns for different goals
Creative combinationsMetaReview enhancement settings, supply real variety
Conversion signalYouPixel plus Conversions API, accurate purchase values
Offer and priceYouFully yours

Where the budget leaks

The removal of the existing-customer cap is the biggest change for D2C brands. Industry reports say Meta cited an EMEA study where campaigns without the cap reached a 13% lower cost per purchase. We could not confirm this from a Meta source. A lower cost per purchase is easy to reach when many of those buyers would have bought anyway.

Watch your new-customer share

Tag every store order as new or returning and compare the share each week. If returning-customer share rises at the same time as platform ROAS, the system may be harvesting demand you already had. Use customer-list exclusions or a separate retention campaign to keep prospecting spend honest.

Creative is now how you target

With audience settings mostly automated, the creative tells the system who the ad is for. A hook built for first-time runners finds first-time runners. Meta AI chat signals sharpen this matching. Meta says sensitive topics such as health, religion, politics and sexual orientation from those chats are not used for ads.

Treat performance claims with care

Several industry guides say Meta's help centre cites a 9% average improvement in cost per conversion for Advantage+ sales campaigns. We could not check that page directly. Some vendors quote much bigger numbers that we could not trace to Meta. Either way, an average cost-per-conversion figure tells you nothing about your margin.

Google and ChatGPT: the new AI ad surfaces

Google now serves your existing ads inside AI Overviews and AI Mode automatically, and ChatGPT has become a self-serve ad channel. Neither gives you full targeting control or separate reporting, so plan your tests around that.

Google: AI Overviews, AI Mode and AI Max

Existing Search, Shopping and Performance Max (PMax) campaigns are automatically eligible for ads in AI Overviews and AI Mode, with no new campaign or budget. Google's help centre confirms you cannot target only these surfaces or opt out of them separately from Search.

To show in AI Overviews, a campaign needs AI-powered targeting: broad match, AI Max keywordless targeting, PMax, Shopping or Dynamic Search Ads. These ads are reported as top ads, so you cannot see their results on their own. AI Overviews has more than 2 billion monthly users in over 200 countries and territories, so this is not a side test.

AI Max is an optimisation layer on top of existing Search campaigns. It includes final URL expansion and gives you more controls than PMax. Google says advertisers typically see 14% more conversions or conversion value at a similar CPA (cost per acquisition) or ROAS (return on ad spend), rising to 27% for campaigns mostly using exact and phrase match. Treat these as Google's averages to test, not results to expect.

Since September 2026, Google has been auto-upgrading text-customisation campaigns to AI Max. Search Engine Land reports that campaigns using the campaign-level broad match setting have also been auto-upgraded from 1 September 2026. Broad match campaigns remain eligible for AI Overviews without AI Max. Newer features, though, such as the Direct Offers asset, only serve with AI Max or text customisation. Your product feed and landing pages now act as targeting, because the system reads them to decide where you appear.

ChatGPT Ads: worth a controlled test

ChatGPT ads began in the US on 9 February 2026 for logged-in adult Free and Go users. On 5 May 2026 OpenAI launched a beta self-serve Ads Manager with cost-per-click (CPC) bidding, after a pilot priced only on cost per thousand impressions (CPM). The minimum spend was removed at the same time.

The rollout was fast. The UK came in August 2026, then India, the Middle East, North Africa and parts of Europe from August to September, depending on the source. South-East Asia followed, taking the total past 60 countries by 23 September.

Pricing has moved just as quickly. Digiday reported CPMs falling by more than half between launch and April 2026. OpenAI's help centre publishes suggested starting CPC bids, so check the figure for your market inside Ads Manager rather than trusting older articles.

Controls are thin. OpenAI's system makes all delivery decisions, and advertisers receive only aggregate metrics, never chat content. For measurement, TechCrunch reports that OpenAI has partnered with Haus, Measured and WorkMagic for geo experiments. It also announced measurement and brand-suitability pilots with DoubleVerify and IAS. Industry reports say Shopify and HubSpot integrations arrived in September 2026. We could not confirm this from OpenAI.

Who should test ChatGPT Ads first

Brands whose buyers research by asking questions: skincare routines, supplements, furniture, financial products, clinics and other service bookings. Impulse products bought from a scroll have less reason to start here.

What changes for brands, agencies and the platforms

Brands now compete on inputs, agencies on strategy and measurement, and platforms on owning every decision. Each shift changes where you should put time and money this quarter.

For brands: your inputs are the controls

Trust matters too. A study cited by BCG found 69% of consumers feel manipulated when brands use AI in ads without disclosing it. If you run AI-generated creative at scale, decide your disclosure policy before a customer asks.

For agencies: from tuning settings to thinking

WPP's then-CEO Mark Read said AI is 'totally disrupting our business'. Dentsu, Omnicom, Publicis and WPP are now partners for buying ChatGPT ads. When bids and audiences are automated, an agency earns its fee through strategy, creative systems, data plumbing and incrementality testing.

Questions to ask any agency that sells 'AI':

If you are weighing options, compare an AI Meta Ads tool against an agency at ₹10 lakh+ a month on cost, speed and control.

For platforms: ads pay for AI

Ads are becoming how AI assistants make money. Meta, Google and OpenAI now compete for the same budget, and each will claim credit for the same sale. Independent checks of brand safety and measurement, from firms such as DoubleVerify and IAS, are becoming the norm, and you should ask for them.

How to measure real profit when AI runs the account

Judge AI campaigns on net delivered revenue against your break-even ROAS, then confirm the result with a holdout test. Platform ROAS is a starting point, not proof.

Even the experts are unsure. Snowflake, citing a 2026 IAB survey, reports that 60-75% of advanced-measurement users said current approaches fall short on rigour, timeliness, trust or efficiency. Platform revenue also usually includes tax, cancelled orders and buyers who would have purchased anyway. For the wider picture, read our guide to attribution after cookies and iOS.

Our order of checks before scaling

  1. Signal: Pixel and CAPI are deduplicated, and purchase values are accurate.
  2. Learning: the ad set has exited the learning phase. Do not judge it before then.
  3. Profit: net delivered ROAS clears your break-even ROAS.
  4. New customers: new-customer share is stable or rising.
  5. Incrementality: a geo holdout, run at least once a quarter, confirms the lift.

Our rule of thumb: raise a budget by no more than 20% every 3-4 days, and only when net delivered ROAS has stayed at least 1.2 times break-even for 7 days. ScaleWin applies the same logic. Before recommending a change, its assistant pulls the last 7-10 days and breaks results down by placement, region, age, gender, hour, ad set and ad. It also checks learning status and compares landing-page views with link clicks to spot page drop-off.

Worked example: a ₹30 lakh a month brand selling in India

Step 1: Start from what Meta reports

Example: a brand spends ₹30,00,000 a month on Meta and sees ₹90,00,000 in reported revenue, a ROAS of 3.0. Average order value is ₹2,000, so that is 4,500 orders. Half are cash on delivery (COD), and 25% of COD orders end as RTO (return to origin: refused or undelivered). Order values include 18% GST, and 35% of buyers are returning customers.

Step 2: Strip out tax and failed deliveries

Remove GST first. Then remove RTO losses: 25% of the 50% COD share means 12.5% of all orders never turn into revenue.

Net delivered ROAS
₹90,00,000 ÷ 1.18 = ₹76,27,000 net of GST. ₹76,27,000 × 0.875 = ₹66,74,000 delivered. ₹66,74,000 ÷ ₹30,00,000 spend = 2.22
Step 3: Compare with break-even and new-customer revenue

Assume a 50% contribution margin after product cost, two-way shipping on RTO orders and payment fees. Break-even ROAS is then 1 ÷ 0.5 = 2.0. At 2.22, the account clears break-even, but only just. New-customer revenue is ₹66,74,000 × 0.65 = ₹43,38,000, a ROAS of 1.45. So first orders from new customers lose money. That only works if repeat purchases pay it back. The headline 3.0 hides all of this.

In COD markets, send purchase events or values that reflect delivered, prepaid or verified orders wherever your setup allows. That way the AI learns from real buyers, not from orders that come back.

Market-by-market notes

What most guides get wrong about AI ad automation

Many top-ranking pages still describe 2024. Here are the claims we see most often, and what is true now.

A 90-day plan for brands spending ₹10 lakh+ a month

Spend the first month fixing signals and settings, the second building creative and tests, and the third trying new channels with proper controls. With BFCM weeks away, avoid starting holdout tests during peak weeks such as Diwali, Ramadan or BFCM, when CPMs swing.

  1. Days 1-15, audit signals: confirm Pixel and CAPI deduplication, check that purchase values are accurate, and tag store orders as new or returning.
  2. Days 1-30, check the AI state: list which Meta campaigns show Advantage+ on or off and why. Find any Google campaigns auto-upgraded to AI Max, and review which pages final URL expansion can send traffic to. Clean the product feed.
  3. Days 31-60, build creative variety: ship new concepts with different hooks, formats and people, not just colour swaps of one ad.
  4. Days 31-60, test AI Max properly: run it as an experiment against a control on your largest Search campaign, and judge on net revenue.
  5. Days 31-60, set up a geo holdout: pause or reduce spend in matched regions to measure true lift.
  6. Days 61-90, test ChatGPT Ads: our rule of thumb is 3-5% of monthly spend with a geo split. On a ₹30 lakh account that is ₹90,000 to ₹1.5 lakh.
  7. Days 61-90, review scope: decide which work your agency, tool or team should own now that the platforms handle bidding and targeting.

At the end of 90 days, you should know your net delivered ROAS, your new-customer share and your true lift. Those three numbers tell you how much to trust the AI, and where to keep your hands on the controls.

Key takeaways

  • Automation is now the default. Meta's Advantage+ and Google's AI Max decide audience, placement and budget split unless you narrow them.
  • New AI ad surfaces are live. You cannot exclude Google AI Overviews and AI Mode on their own, and ChatGPT Ads are available in more than 60 countries.
  • Meta removed the cap on spend to existing customers, so you must track new-customer revenue yourself.
  • Judge AI-run campaigns on net delivered ROAS after tax, returns and RTO, not on platform ROAS.
  • Test ChatGPT Ads with a small, geo-controlled budget before you move real money.

Ai and advertising: frequently asked questions

How is AI used in advertising today?

Ad platforms use machine learning to choose audiences, placements, bids and budget splits, and generative AI to create or combine ad variations. On Meta and Google this is now the default setting. ChatGPT Ads also uses AI to decide when and where ads appear.

Will AI replace advertising agencies?

It is replacing much of the manual work of setting up and tuning campaigns. Agencies that add strategy, creative systems, data setup and incrementality testing still add value. Agencies that only adjust bids and audiences are competing with features the platforms give away free.

Can I opt out of ads in Google AI Overviews or AI Mode?

No. Google says you cannot target only AI Overviews or AI Mode, and you cannot opt out of them separately from Search. Eligible Search, Shopping and Performance Max campaigns serve there automatically.

Is AI Max required to show ads in AI Overviews?

No. Broad match Search, Performance Max, Shopping and Dynamic Search Ads campaigns are eligible without AI Max. Some newer features, such as the Direct Offers asset, only serve with AI Max or text customisation.

Can I still limit Advantage+ spend on existing customers?

Not with a percentage cap. Meta has removed that field. You can use customer-list exclusions or separate campaigns, and you should track new versus returning customer revenue in your store every week.

Does Meta use my Meta AI chats to target ads?

Since 16 December 2025, Meta uses interactions with its AI features to personalise ads, except in the UK, the EU and South Korea. Meta says sensitive topics such as health, religion and politics are not used for ads.

Are ChatGPT ads available in India, the UAE, the UK or Australia?

ChatGPT Ads launched in the UK in August 2026, and in India and parts of the Middle East from August to September 2026. They were in more than 60 countries by late September 2026. OpenAI says ChatGPT Ads has also launched in Australia, but check availability for any specific country inside Ads Manager.

How much do ChatGPT ads cost?

Costs vary by market and are changing fast. Digiday reported CPMs falling by more than half between launch and April 2026, and CPC bidding arrived in May 2026. OpenAI publishes suggested starting bids, so check the current figure for your market in Ads Manager.

How do I measure whether AI-run campaigns are actually incremental?

Run a geo holdout, where you pause or cut spend in matched regions and compare sales against regions that keep running ads. Pair it with net delivered ROAS and new-customer revenue from your store, not platform-reported ROAS alone.

Do I pay GST on Meta and Google ads in India, and does the equalisation levy still apply?

Paying foreign ad platforms is treated as importing a service, so 18% GST applies under reverse charge. Some advertisers are billed by Indian entities with GST on the invoice, so check yours. The 6% equalisation levy on online ads was abolished from 1 April 2025.

Sources

Ajay Kumar
Founder, ScaleWin · 12+ years in performance marketing

Ajay is the founder of ScaleWin. He works hands-on with Meta Ads, Google Ads and SEO for Indian D2C and service brands, and built ScaleWin so that the daily work of a performance marketer - reading numbers, pausing losers, scaling winners - happens automatically.

Drafted with AI assistance. How these articles are written

Let ScaleWin read your ad data every day

ScaleWin checks your Meta and Google Ads daily, judges every ad on profit and tells you exactly what to change.

Start free, no card

Get marketing tips that grow your business

One practical email a week on ads, SEO and growth. No spam, no fluff. Unsubscribe anytime.