AI Sparks are insight cards that surface the handful of anomalies and opportunities across your accounts worth your attention. Nothing a Spark flags ever changes a budget or touches a campaign.
Think of a Spark as a note from a senior analyst who watched your accounts overnight. Each one carries a short headline, a one-sentence diagnosis, a specific recommendation, and a colour-coded severity badge (info, success, warning, or critical). Sparks read across all five platforms: Google Ads, Meta Ads, LinkedIn Ads, Microsoft Ads, and TikTok Ads.
Where you'll find them
Sparks show up in three places:
- Recent Sparks on your Home page.
- The Sparks popover, opened from the sparkles icon in the top bar (not the notification Bell).
- The account-level Sparks tab, with full history and search.
New Sparks can also reach you through Slack, email, and the Bell. For a fuller map of where these sit, see understanding the Pace dashboard.
What Sparks flag
A Spark fires only when something is significant. Roughly 90% of the time the analysis returns nothing at all, which is what keeps the feed worth reading. When one does fire, it's usually one of:
- Cost and results anomalies: CPA (cost per acquisition) climbing week over week, ROAS (return on ad spend) sliding, a conversion campaign spending with zero conversions, or conversion volume dropping while spend holds steady.
- Positive signals: conversions up while cost holds or improves, or an ad resonating on click-through.
- Platform-specific issues: a single Google Ads keyword eating budget with no conversions, Meta Ads creative fatigue, or LinkedIn Ads lead-form drop-off.
Sparks inform but don't act
A Spark's recommendation is a sentence in the card, not an action Pace takes on your behalf. Sparks never pause a campaign, move a budget, or write anything back to the ad platform. That budget-changing automation is a separate system — read how Pace decides budget changes for that side of the product, or for a broader look at where AI genuinely helps in ad management versus where it's just relabelled automation, see this breakdown.
Why the feed stays calm
Sparks avoid repeating themselves. When one flags a problem, you hear about it once, then it goes quiet for three days even if the issue persists. If it's still live after that, Sparks re-flag it once so you know it hasn't been resolved. Pace tracks this quiet period per account and per platform, so a noisy issue on one account never silences the same pattern on another. A brand-new critical finding always comes through regardless.
You stay in control too. Mute a Spark to silence future similar findings for that account. Choose 7 days, 30 days, 90 days (the default), or until you reactivate it. Hide removes just that one card. Muted findings don't vanish: a neutral Muted pill replaces the severity badge, and one click reactivates alerting.
The tone is deliberate. Pace writes Sparks like an analyst leaving a note: plain, factual, and never shouting. Severity lives in the colour-coded badge, so a critical finding still reads as clearly critical without an all-caps prefix.
Note: AI Sparks are available on the Pro plan and above. On Basic, the Recent Sparks panel shows an upgrade prompt instead. Sparks run per account on a schedule you set, so findings arrive on that cadence rather than the instant something changes.