top of page

How to Measure an AI-Generated Campaign: Goals, KPIs and Testing Against Shot Imagery

Oct 30
7 min read

Updated: 2 days ago

To measure an AI-generated campaign, write down the goal and the metric before production starts, then judge the imagery by the same standards you would apply to a photographed campaign. Pick one primary objective, such as awareness, consideration or sales, and assign each stage its own key figure: reach and recall for awareness, click-through and engaged visits for consideration, conversion rate and revenue for sales. If you want to know whether AI imagery performs as well as shot imagery, run a controlled test. Show both versions to comparable audiences at the same time, with identical copy, placement, budget and offer, so that the image is the only difference. Agree on the sample size and the test duration in advance and do not stop early when one version leads. Some effects, such as long-term brand perception, need surveys or brand-lift studies rather than click data. Production speed and cost per usable asset belong in the evaluation as well, but in a separate column.


Set the objective before the studio starts


Most measurement problems start before the first image exists. A brand commissions AI imagery because it is new, the campaign runs, and afterwards the team looks for numbers that justify the decision. You can always find such numbers, and they prove very little.


Decide three things before the brief goes to the studio:


  • The business objective. Is the campaign meant to make a new collection known, to move people from interest to the product page, or to sell a defined range within a defined period? One campaign can serve several stages, but one of them has to lead.

  • The primary metric. One number that tells you whether the campaign met the objective.

  • The comparison point. A previous campaign, a control group, a parallel photographed version or a market where the campaign does not run. Without a baseline, a figure has no meaning.


Put these points into the brief. A studio that knows the campaign has to win a click test on social placements will make different choices from one producing a hero image for print. What to Put in a Brief for an AI Production Studio lists the other parts of the brief.


KPIs by funnel stage


The right KPI depends on the job the image has to do.


Awareness


The image has to be seen and remembered. Typical figures are reach, frequency, video completion for motion assets, and ad recall measured through surveys. Budget and targeting drive reach, so it says little about the image. Recall and recognition tell you more.


Consideration


The image has to make people want to look closer. Click-through rate, engaged sessions on the landing page, time on the product page and saves or shares are common measures. Click-through rate reacts directly to the visual, which makes it a sensible primary metric for a creative test.


Conversion


The image has to support a purchase or an enquiry. Conversion rate, revenue per visitor, average order value and, for some categories, the return rate matter here. Returns deserve attention in fashion and accessories. If generated imagery makes a fabric look softer or a colour warmer than the real product, customers notice when the parcel arrives, and the cost appears weeks after the click.


Brand


Premium and luxury brands care about how a campaign shapes perception over time. Click data does not capture this. Brand-lift studies, panel surveys and comment analysis give a better picture: do people describe the brand in the terms you want, and do they notice anything that feels artificial?


How to test AI imagery against shot imagery


Quick comparisons between a generated and a photographed image rarely tell you anything, because too much else differs between them.


Change one thing only


Both versions must share everything except the image: the same headline, the same call to action, the same offer, the same landing page, the same placement and the same budget. If the AI version also gets a new headline, you no longer know what caused the difference.


Keep the content of the images comparable too. Pitting a generated outdoor scene against a photographed studio packshot tests the setting, not the production method. The fair test shows the same product, in a similar composition, with the same styling direction, once produced with a camera and once generated or composited.


Split the audience properly


Run both versions at the same time, to randomly assigned, comparable audiences. Sequential tests, where the shot version runs in one month and the AI version in the next, mix the image effect with season, competition and news. Most major ad platforms offer a split-testing function that divides the audience at random. Use it rather than two separate campaigns that compete for the same people.


Decide sample size and duration in advance


Before the test starts, decide the smallest difference that would change your decision, calculate the sample size for that difference with a standard calculator, and fix the duration. The lower your baseline rate and the smaller the difference you care about, the larger the audience has to be. A full week or more usually smooths out weekday patterns. If you stop the test because one version looks ahead after two days, you will often crown a false winner.


Read the result with care


A result that is not statistically clear means the images performed alike within the precision of the test. For many brands that outcome is useful, because it means AI imagery can carry the placement at a different cost and speed. Also check the result per segment and per placement. An image can win in a feed and lose in a vertical story.


What you can measure and what you cannot


Clicks, conversions, cost per result and production time are easy to count. Three things take more effort:


  • Long-term brand effects. A single campaign rarely moves brand perception measurably on its own. Surveys before and after, or a brand-lift study, give an indication.

  • Audience reaction to the use of AI itself. Disclosure can change how people respond to an image. Where you label AI content, the label is part of what you test, so test the version you will run.

  • Creative fatigue. AI production makes it easy to rotate many variants. To see whether rotation holds performance up, track results beyond the first week.


Say in the report which questions the data answers and which it leaves open.


Production metrics belong in a separate column


AI and hybrid production change how a campaign gets made: time from brief to first draft, number of correction rounds, number of approved assets, cost per usable asset. These figures count, because more market and format variants in less time give a brand flexibility.


Keep them apart from media performance. "Cheaper to produce" and "performs better" are two separate claims, and a single blended number hides weak results. Use a scorecard with two columns: how the campaign performed in market, and what it cost in time and money to make. The comparison with traditional production in Hybrid Campaigns vs. Traditional Production shows where those differences come from.


A simple measurement plan


Before the next AI or hybrid campaign, write one page that covers:


1. The objective and the primary metric.


2. The baseline or control.


3. The test design: what varies, what stays the same, audience split.


4. Sample size, duration and the rule for when the test ends.


5. Secondary metrics per funnel stage, including returns where relevant.


6. Brand measures, if perception matters.


7. Production metrics, kept in a separate column.


8. Who reads the results and which decision depends on them.


Send this page to the studio and the agency together with the brief, so all three parties judge the campaign by the same rules.


Frequently Asked Questions


How do you measure the success of an AI-generated campaign?

Set one business objective and one primary metric before production, choose a baseline or control group, and track KPIs per funnel stage: recall for awareness, click-through for consideration, conversion and revenue for sales. Report production time and cost separately from media performance.


How do AI-generated ads compare with traditionally produced campaigns?

There is no general answer, because the result depends on the brand, the product and the placement. The reliable way to find out is a split test that shows a generated and a photographed version of the same product to comparable audiences at the same time, with copy, offer and budget held identical.


What KPIs should a brand use for an AI campaign?

The same KPIs as for any campaign, chosen by objective: reach, frequency and ad recall for awareness; click-through rate and engaged visits for consideration; conversion rate, revenue per visitor and return rate for sales. Brand-lift studies or surveys cover perception.


How long should an A/B test of AI imagery run?

Long enough to reach the sample size you calculated in advance, and at least a full week in most cases, so that weekday patterns even out.


Should production cost count as a success metric?

Yes, but as a separate figure. Cost per usable asset and time from brief to approval show the production effect. Media performance shows the market effect.


Can you measure how audiences react to AI imagery being disclosed?

Partly. If you label AI content, test the labelled version you will run, so the result includes the label's effect.


Planning a campaign with AI imagery or video? Request an AI campaign.


About The Bridges AI

The Bridges AI is a hybrid AI creative agency: a creative studio and production partner for fashion, beauty, luxury and hospitality brands and for the agencies that work with them. It combines booked real talent, real products and human creative direction with AI production, and offers licensed digital twins of real models, only of models under contract. The Bridges AI works in strategic partnership with Most Wanted Models®, the German model agency founded in 1997, which brings nearly 30 years of experience in bookings, buyouts and usage rights.


Written by the editorial team of The Bridges AI. Last updated October 2026.


Comments


bottom of page