AD FIELD NOTESAI advertising · the work behind the output

Case files / ai assisted copy testing

Pandora: make AI-assisted ad copy feel like a personal charm

Pandora and Microsoft Advertising used Copilot, inclusive language and preference testing to make search copy feel more emotionally specific—showing an AI advertising workflow that changes the message, not the product image.

Campaign period: 8 July 2026Source published: 8 July 2026Archive entry prepared: 15 September 2026

Retrospective coverage, prepared in September 2026 from the original record.

Pandora jewelry charms photographed for a Microsoft Advertising case study about emotionally intelligent ad copy.
Pandora’s charms frame the case study’s argument for more personal advertising language. Pandora / Microsoft Advertising; image published in Microsoft’s customer story.

Reported execution: variation was the creative unit

Microsoft describes Pandora’s test as an effort to make every shopper feel represented in paid and organic search. The team developed ad variations with inclusive emotional modifiers, then used Copilot in Microsoft Advertising to create and test options that stayed within Pandora’s voice.

A choice-based survey complemented the platform work. The campaign was not simply an automated copy swap: audience targeting, search structure and human interpretation were part of the same loop.

Our reading: emotional specificity is a better brief than ‘personalization’

The useful move is to define personalization as a feeling the message should make possible—being seen, celebrated or included—rather than as a vague promise to use more data. That gives the team something people can actually compare.

It also leaves room for disagreement. Preference testing can show which line resonates, but the brand still has to decide whether the line is true to its product and audience instead of selecting the highest score by reflex.

Production lesson: test the reason to believe, not only the wording

A copy-variation system should preserve the hypothesis behind each modifier, the audience it is meant to help and the destination experience it points toward. Otherwise a high-performing phrase becomes a reusable trick detached from the brand promise.

For teams using generative copy tools, a small message taxonomy and a human tone review are inexpensive guardrails. They make it possible to learn from the test without turning inclusive language into a decorative layer.

Evidence boundary: platform case studies are directional

Microsoft reports that overall purchase intent rose from 70% to 77%, non-owner intent from 46% to 64%, and EQ-led ads earned a 40% higher Preference score than generics. The case page does not publish the full sample, experiment design or independent replication.

The record supports a concrete workflow—AI-assisted variations plus preference testing—not a universal claim that emotionally intelligent copy will produce the same lift in another category or market.

Sources & limits

Microsoft Advertising customer story dated 8 July 2026; purchase-intent and preference figures are Microsoft/Pandora case claims, not an independent effectiveness audit.

  1. Microsoft Advertising — Pandora creates ads as personalized as custom charms — 8 July 2026.

Send a correction with the passage and supporting source.

The weekly field note

One useful note for the next brief.

Campaigns worth studying, the work behind them, and questions to bring into your next creative review. Delivered weekly.

How we handle your email