The change in a brand's visibility when AI search systems apply different weighting or ranking signals.
Readers use it to measure how changes like schema markup or content updates impact a brand's appearance in AI-generated results.
01what it is and how it works
Treatment Effect is calculated by comparing visibility scores before and after a change in how AI ranking models treat the brand. The system records the number of times the brand appears in AI-generated answers, then applies a statistical difference to isolate the impact of the treatment, such as a new schema markup or a content update. The result is a numeric effect that can be positive or negative.
It shows how much a brand's visibility changes when AI search treats it differently.
02what to do about it
To address a negative Treatment Effect, first verify the change you made is correctly implemented, then monitor visibility daily for at least three days. If the effect is small, consider adjusting the treatment or testing alternative signals. Publish updated content, add relevant structured data, and re-run the measurement to confirm improvement.
03how it is measured or noticed
The effect is measured by tracking the frequency of the brand's mentions in AI-generated search responses over time. Compare the count before and after the treatment using the product's analytics dashboard, which shows impressions, click-through rates, and ranking positions. A sustained shift in these metrics indicates a real Treatment Effect.
04common mistakes
- Assuming correlation equals causation without a control group
- Relying on a single metric like impressions instead of multiple signals
- Ignoring time decay in AI ranking changes
- Failing to update tracking after schema changes
05limits
This metric only applies when AI search actively surfaces the brand in generated answers; it is irrelevant for pure text results or non-AI platforms. It can be confused with overall SEO ranking changes, which are measured by traditional search engines rather than AI output. Use it only for AI-driven visibility shifts.
06worked example
For example, after adding FAQ schema, the brand's AI answer appearances rose from 120 to 138 per week. The product logged a 15% increase, which the dashboard highlighted as a positive Treatment Effect.
Rich results can increase the visibility of a page in Search.
Frequently asked questions
How is Treatment Effect different from overall brand visibility?
Treatment Effect isolates the change in visibility caused by a specific intervention, while overall visibility is the total brand presence in AI-generated answers. It compares visibility scores before and after a change in how AI ranking models treat the brand, so it reflects cause and effect rather than a static snapshot. Overall visibility can shift for many reasons, but Treatment Effect attributes the movement to a deliberate treatment.
Should I act on a negative Treatment Effect, and what determines that decision?
Yes, if the drop is large enough to matter for your business goals, you should investigate and act. The decision depends on how much the change affected your brand's appearances in AI answers and whether the treatment was intentional or accidental. A small dip may be noise, but a sustained decline after a known change usually warrants correction.
How exactly is Treatment Effect calculated, and who can run it?
It is calculated by comparing the brand's visibility scores before and after a change in how AI ranking models treat the brand, using tracked mention frequency in AI-generated responses. The system that measures brand appearance in AI search performs this comparison automatically when a treatment is applied. Marketing and SEO teams can review the resulting scores, but the calculation itself is handled by the measurement platform.
Does Treatment Effect still work if AI search stops showing my brand?
No, Treatment Effect only applies when AI search actively surfaces the brand in generated answers. If the brand no longer appears in AI-generated results, there is no visibility to measure before or after a treatment. It is irrelevant for pure text results or non-AI platforms, so the metric cannot capture changes outside of AI-generated answer surfaces.
What breaks if I ignore a negative Treatment Effect, and how would I notice?
Ignoring it can lead to sustained loss of presence in AI-generated answers, which may reduce referral traffic and brand discovery over time. You would notice by tracking the frequency of the brand's mentions in AI-generated search responses and seeing a continued downward trend. The cost of getting it wrong is a gradual erosion of visibility that becomes harder to recover once competitors fill the gap.
How long before a Treatment Effect shows up, and what should I measure meanwhile?
Effects often appear within days, but you should monitor visibility daily for at least three days to confirm a trend. Meanwhile, track the frequency of the brand's mentions in AI-generated search responses to catch early signals. Waiting longer than a week without checking risks missing a sharp drop or a false recovery.
Asked out loud
spoken, not typedThe same term in the words somebody uses speaking to an assistant rather than typing into a box — written from the situation, which is why each one carries the situation it came from.
It depends on whether the change is still rolling out across AI search systems. If appearances are still falling after a full day, you should verify the schema implementation is correct and monitor daily for at least three days. Letting it run unchecked risks a sustained drop in visibility.
It depends on whether the change reduced how often AI models surface the brand in generated answers. If appearances dropped significantly and haven't recovered, you likely need to adjust the content or markup. Check the frequency of mentions in AI-generated responses over the next few days to confirm the trend.
Usually, if the dip follows a recent change you made, you should verify it is correctly implemented before deciding. Monitor visibility daily for at least three days to see if it recovers on its own. If it keeps falling, rolling back the change is safer than waiting.