AI advertising · Glossary

Advertising and AI Glossary: 40 Terms Every Marketer Should Know

Forty terms you will meet in ad accounts, agency reports and platform updates, each with what it means and why you should care.

Ajay Kumar • •10 min read
Illustration of an ad card with a label, for a glossary of AI advertising terms

Quick answer: which AI advertising terms matter most?

The terms that matter most in AI advertising are the ones that change your decisions: the campaign types that automate delivery, the learning phase, the data signals the system learns from, and the measures of real return such as break-even ROAS and blended return. This glossary explains 40 of them in plain words.

How to use this glossary

This page is a reference, not an article to read top to bottom. It groups 40 terms into four sets, and each term has a plain meaning and a reason to care.

  1. Find the set. Campaign types, bidding and delivery, data and measurement, or costs and compliance.
  2. Read the third column first. It tells you what the term changes in your decisions.
  3. Follow the link in each section when you need the full method.

Where a meaning comes from a platform's own page, we link to it. Where a number is a platform's own guidance, we say so.

If you want the reasoning behind these terms in one place, start with our founder's guide to advertising and AI.

AI campaign types and platform terms

These are the product names you will see when a platform offers to automate a campaign. They differ by platform, but most automate the same jobs.

TermPlain meaningWhy it matters
Advantage+ sales campaignMeta's automated sales campaign. AI chooses audience, placement and budget split.You steer it with creative and conversion data, not with targeting.
Advantage+ audienceMeta's setting that lets the system look beyond the audience you suggest.Your audience choice becomes a hint, not a limit.
Performance MaxOne Google campaign that runs across Search, YouTube, Display, Discover, Gmail and Maps.It depends on your assets, product feed and conversion values.
AI Max for SearchGoogle's AI features for Search: search term matching, text customisation and final URL expansion.It can choose the searches, the ad text and the landing page.
Dynamic Search AdsOlder Google Search ads built from your website content.Google announced these would be upgraded to AI Max from September 2026.
Demand GenGoogle campaigns for YouTube, Shorts, Discover, Gmail, Maps and the Display Network.Display campaigns are moving into it during 2026.
Performance+Amazon's AI campaign type for video and display.It works towards a sales or acquisition goal, on and off Amazon.
Smart BiddingGoogle's automated bidding, set for each auction in real time.You give a goal. The system sets the bid.
Responsive search adA search ad where you supply several headlines and descriptions and Google mixes them.Google allows up to 15 headlines and 4 descriptions.
Product feedThe file of product titles, images, prices and stock sent to the ad platform.In shopping and Performance Max campaigns, the feed is the ad.

Google's limits for responsive search ads are on its help page, and the AI Max change is in Google's April 2026 announcement. For how these products compare, see AI advertising across channels.

Bidding and delivery terms

These terms describe how the system decides who sees your ad and what you pay. They explain most sudden changes in results.

TermPlain meaningWhy it matters
AuctionThe contest that decides which ad is shown each time a slot is available.You pay based on competition, not a fixed price list.
Bid strategyThe rule you give the system for bidding, such as lowest cost or a target.Changing it can restart learning.
Target ROASA bid strategy that aims for a set return on ad spend.Set it from your break-even, not from habit.
Target CPAA bid strategy that aims for a set cost per order or lead.Too low a target and the campaign stops spending.
Learning phaseThe period when the system is still working out how to deliver an ad set.Meta says it usually ends after about 50 results in a week.
Learning LimitedMeta's status for an ad set that is not getting enough results to finish learning.Waiting does not fix it. More results or fewer ad sets do.
Significant editA change that sends an ad set back into learning.Batch your edits into one day a week.
FrequencyThe average number of times each person saw your ad.A rising number with falling results points to tired creative.
PlacementWhere the ad appears, such as Feed, Reels, Stories or Search.AI campaigns choose placements for you.
Creative fatigueMeta's status when an ad costs at least twice as much per result as your past ads.It confirms a decline you should already have seen.

Meta's wording on the learning phase is on its help page. Our guide to measuring profit shows how to set a target from your own costs.

Data and measurement terms

AI campaigns learn from the data you send back. These terms cover that data and the reports built on it.

TermPlain meaningWhy it matters
PixelCode on your site that reports visitor actions to the ad platform from the browser.It misses some events, so it should not work alone.
Conversions API (CAPI)A way to send the same events from your server.Meta recommends using it together with the pixel.
DeduplicationKeeping one copy when the browser and the server report the same event.Without it, sales are counted twice.
Event match qualityMeta's score from 0 to 10 for how well your events match to accounts.Better matching gives the system more to learn from.
Enhanced conversionsGoogle's feature that sends hashed customer data with your conversions.It improves measurement when cookies are limited.
Attribution windowHow long after a click or view a sale is still credited to the ad.Change the window and the same campaign reports a different result.
Engage-through attributionMeta's 2026 name for conversions after a like, share, save or video view.These no longer count as click-through conversions.
IncrementalityThe sales that happened only because of the ads.It is measured with a lift test, not from a normal report.
ROASReturn on ad spend: reported revenue divided by ad spend.Useful to compare ads. Weak as proof of profit.
Blended return (MER)Total store revenue divided by total ad spend, all channels.Nothing is counted twice, so it is the honest overall view.

Three of these terms in one sum

ROAS, blended return and cost per order are often mixed up. A short example shows how they differ.

Worked example: a brand spending ₹24 lakh a month

Step 1: Platform ROAS

Example: Meta spend is ₹20,00,000 and Meta reports ₹70,00,000 of revenue.

ROAS
₹70,00,000 ÷ ₹20,00,000 = 3.5x
Step 2: Blended return

Google spend is another ₹4,00,000. The store took ₹96,00,000 in total that month.

Blended return
₹96,00,000 ÷ ₹24,00,000 = 4.0x
Step 3: Cost per order

Meta reports 2,500 orders, so each cost ₹20,00,000 ÷ 2,500 = ₹800. If the brand can afford ₹900 an order, Meta is inside the limit.

The full signal checks are in what AI needs from your pixel, CAPI and store.

Cost, profit and compliance terms

The last set covers the money terms every report uses and the disclosure terms that arrived with AI-made ads.

TermPlain meaningWhy it matters
CPMCost per 1,000 impressions.It shows how expensive attention is, not whether it sold anything.
CTRClick-through rate: clicks divided by impressions.A sign of how well the creative earns attention.
CPACost per acquisition: ad spend divided by orders or leads.Compare it with what you can afford per order.
Break-even ROASThe ROAS at which profit after ads is zero.Below it you lose money, whatever the dashboard colour.
RTOReturn to origin: a parcel refused or not delivered, sent back to you.Common with cash on delivery. It was already counted as a sale.
AI info labelMeta's label on ads made or heavily edited with generative AI.It is applied automatically in many cases.
C2PAA technical standard that records how an image or video was made.Meta uses it to detect content from outside AI tools.
Synthetic performerAn AI-generated person in an ad who appears real.New York law requires ads to disclose them.
DeepfakeAI-made or AI-changed image, audio or video that looks real.The EU requires disclosure from 2 August 2026.
UGCUser-generated content: ads filmed in a customer's style.Paid creators and AI versions both need honest disclosure.

Meta explains its labels on its AI info help page, and the EU rule is described in the European Commission's guidelines on Article 50. For when to label an ad, see our guide to AI advertising disclosure rules.

One question for any new term

When a report or a sales call uses a term you do not know, ask: how is it calculated, and from which data? If nobody can answer, do not base a budget decision on it.

What most glossaries get wrong about AI advertising terms

Most glossaries copy the platform's definition and stop. The trouble starts where two terms sound alike and mean different things.

Terms that are new or changed in 2026

If your team learned these terms before 2026, some meanings have moved. These are the ones to update.

Check old reports

A report template made in 2025 may still show columns that no longer fill, or compare numbers counted in two different ways. Review your templates once.

Money terms behave differently by market too. Order value often includes GST in India or VAT in the UK, EU and Gulf, and RTO matters wherever cash on delivery is common. ScaleWin applies these corrections for you when it reports profit for each ad.

Key takeaways

  • AI campaign names differ by platform, but they automate the same three things: audience, placement and bidding.
  • Learning phase, significant edit and Learning Limited explain most "why did results change" questions.
  • ROAS, blended return and break-even ROAS are three different numbers. Only the last one tells you if you make money.
  • Several terms are new or changed in 2026: AI Max, engage-through attribution, AI info labels and synthetic performers.
  • When a report uses a term you do not know, ask how it is calculated before you act on it.

Advertising and ai: frequently asked questions

What is the difference between ROAS and blended return?

ROAS is revenue reported by one ad platform divided by spend on that platform. Blended return, also called MER, is total store revenue divided by total ad spend across all channels. Platforms can claim the same sale, so blended return is the more honest overall number.

What does learning phase mean in Meta ads?

It is the period when Meta's system is still working out how best to deliver an ad set. Meta says it usually ends after about 50 results in the week after the last significant edit. Results are less stable during this time, so avoid judging or editing too early.

What is the difference between Performance Max and AI Max?

Performance Max is a Google campaign type that runs across all Google inventory from one campaign. AI Max is a set of AI features inside Search campaigns: search term matching, text customisation and final URL expansion. You can use both in the same account.

What is engage-through attribution?

It is Meta's name, introduced in 2026, for conversions that follow an interaction other than a link click, such as a like, share, save or video view. Meta now reports these separately, and click-through attribution counts link clicks only.

What is an AI info label on an ad?

It is a label Meta adds to ads that were created or significantly edited with generative AI. Meta applies it automatically when you use its own AI tools, and when it detects standard metadata from outside tools. In some regions, advertisers can also disclose AI use themselves.

What does RTO mean in ecommerce advertising?

RTO means return to origin. It is an order that was refused or could not be delivered and came back to the seller. It is common with cash on delivery. The ad platform already counted it as a sale, so reported revenue is higher than real revenue.

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.