Competitive Ad Intelligence

Data Sources

All ad libraries are public and free to access — this is publicly available competitive intelligence.

Output Schema

const AdIntelligenceSchema = z.object({
  scanDate: z.string(),

  competitorAdActivity: z.array(z.object({
    competitor: z.string(),
    platform: z.string(),
    totalActiveAds: z.number(),
    newAdsSinceLastScan: z.number(),
    estimatedSpendLevel: z.enum(["minimal", "moderate", "significant", "heavy"])
      .describe("Based on ad volume, platforms, and longevity"),

    adBreakdown: z.object({
      formats: z.record(z.string(), z.number())
        .describe("{ 'single_image': 5, 'video': 3, 'carousel': 2 }"),
      objectives: z.array(z.string()),
      themes: z.array(z.string())
        .describe("Main messaging themes: 'holiday sale', 'new product', 'brand awareness'"),
      ctas: z.array(z.string()),
      landingPages: z.array(z.string()),
    }),

    topAds: z.array(z.object({
      adId: z.string().optional(),
      platform: z.string(),
      format: z.string(),
      headline: z.string().optional(),
      bodyText: z.string().max(300),
      cta: z.string(),
      landingPage: z.string().optional(),
      startedAt: z.string().optional(),
      longevity: z.string().describe("How long it's been running — longer = likely performing well"),
      analysis: z.string().describe("What makes this ad effective or notable"),
    })).max(5),

    strategicInsights: z.array(z.string())
      .describe("What their ad strategy reveals about their business priorities"),
  })),

  campaignAlerts: z.array(z.object({
    competitor: z.string(),
    alertType: z.enum([
      "new_campaign_launch",
      "spend_increase",
      "new_messaging_angle",
      "targeting_your_audience",
      "holiday_push",
      "product_launch",
      "competitive_attack",
    ]),
    description: z.string(),
    urgency: z.enum(["fyi", "action_recommended", "urgent"]),
    evidence: z.string(),
  })),

  counterPositioning: z.array(z.object({
    competitorAction: z.string(),
    suggestedResponse: z.string(),
    responseType: z.enum([
      "counter_ad",           // Run your own ad targeting the same audience
      "organic_content",      // Create organic content differentiating from their message
      "ignore",               // Not worth responding to
      "differentiate",        // Highlight what you do that they don't
      "undercut",             // Offer a better deal
      "outflank",             // Target an adjacent audience they're missing
    ]),
    timing: z.string(),
    suggestedAdCopy: z.string().optional(),
    estimatedBudgetNeeded: z.string().optional(),
  })),

  marketOverview: z.object({
    totalCompetitorAds: z.number(),
    mostActiveCompetitor: z.string(),
    dominantFormat: z.string(),
    dominantTheme: z.string(),
    industryAdTrend: z.enum(["increasing", "stable", "decreasing"]),
    whitespace: z.array(z.string())
      .describe("Messaging angles or formats no competitor is using"),
  }),

  confidenceScore: z.number().min(0).max(1),
});

Database

model CompetitorAd {
  id              String   @id @default(uuid())
  organizationId  String
  competitorName  String
  platform        String
  adLibraryId     String?
  format          String
  headline        String?
  bodyText        String?  @db.Text
  cta             String?
  landingPage     String?
  firstSeenAt     DateTime @default(now())
  lastSeenAt      DateTime @default(now())
  isActive        Boolean  @default(true)
  screenshotUrl   String?
  analysis        String?  @db.Text

  @@unique([organizationId, platform, adLibraryId])
  @@index([organizationId, competitorName, isActive])
  @@index([organizationId, firstSeenAt])
}

model AdIntelligenceReport {
  id              String   @id @default(uuid())
  organizationId  String
  organization    Organization @relation(fields: [organizationId], references: [id], onDelete: Cascade)
  scanDate        DateTime
  competitorData  Json
  alerts          Json
  counterPositioning Json
  marketOverview  Json
  createdAt       DateTime @default(now())

  @@index([organizationId, scanDate])
}

Schedule

// Daily: scan ad libraries for changes
export const adIntelligenceScan = inngest.createFunction(
  { id: "ad-intelligence-scan" },
  { cron: "0 3 * * *" },
  async ({ step }) => {
    // 1. For each org's tracked competitors
    // 2. Query Meta Ad Library, TikTok Creative Center, etc.
    // 3. Identify new ads, removed ads, long-running ads
    // 4. Store/update CompetitorAd records
  }
);

// Weekly: deep analysis with LLM
export const adIntelligenceAnalysis = inngest.createFunction(
  { id: "ad-intelligence-analysis" },
  { cron: "0 7 * * 2" },  // Tuesday
  async ({ step }) => {
    // 1. Aggregate weekly ad data
    // 2. Run LLM analysis for insights, alerts, counter-positioning
    // 3. Feed to Strategy Agent, Ad Copy Agent, Reporting Narrator
  }
);

// Real-time: alert on major campaign launches
export const adIntelligenceAlert = inngest.createFunction(
  { id: "ad-intelligence-alert" },
  { event: "competitor/major-ad-activity" },
  async ({ event, step }) => {
    // Triggered when daily scan detects >5 new ads from one competitor
    // or a significant messaging shift
    // Immediate notification to dashboard
  }
);

Integration

Competitive Ad Intelligence Agent
├── counterPositioning → Ad Copy Agent (create counter-ads)
├── counterPositioning → Content Creator (organic counter-content)
├── campaignAlerts → Strategy Agent (adjust strategy in response)
├── campaignAlerts → Pricing Intelligence (cross-reference pricing moves)
├── whitespace → Content Creator + Ad Copy (exploit gaps competitors miss)
├── topAds → A/B Testing (test similar approaches for client)
└── marketOverview → Reporting Narrator (include in competitive section of reports)