Discovery Agent

AI-powered creator discovery and recommendation from 38M+ profiles

Last updated: 2026-01-19

The Discovery Agent continuously searches for creators who match your brand, analyzing content, audience demographics, and engagement patterns across a database of 38M+ profiles. It surfaces the best opportunities so you can build a high-performing affiliate roster.

Learning Over Time

The Discovery Agent learns from your approve/reject decisions. The more feedback you provide, the better its recommendations become. Most programs see significantly improved match quality after 2-3 weeks of active use.

What It Does

The Discovery Agent runs continuously in the background, performing these tasks:

  • Daily Database Scans — Searches 38M+ creator profiles for new matches
  • Content Analysis — Evaluates recent posts for brand alignment and quality
  • Audience Evaluation — Analyzes follower demographics against your customer profile
  • Competitor Monitoring — Identifies creators mentioning your competitors
  • Brand Mention Detection — Finds creators already talking about your brand
  • Lookalike Discovery — Finds creators similar to your top performers
  • Prioritized Recommendations — Generates ranked lists based on fit scores

Configuration Settings

Configure the Discovery Agent to match your ideal affiliate profile. Access settings at AI → Agent Insights → Discovery Agent → Settings.

Follower Range

Set minimum and maximum follower counts to target the right creator tier:

  • Nano (1K-10K) — High engagement, niche audiences, lower reach
  • Micro (10K-100K) — Balanced engagement and reach, often best ROI
  • Mid-tier (100K-500K) — Broader reach, established credibility
  • Macro (500K-1M) — Large audiences, brand awareness plays
  • Mega (1M+) — Celebrity-level reach, premium partnerships

Default: 5,000 - 500,000 followers (optimized for micro and mid-tier)

Platform Preferences

Select which platforms to search. Enable multiple platforms for broader discovery:

  • Instagram — Posts, Reels, Stories analysis
  • TikTok — Video content and engagement patterns
  • YouTube — Long-form content, tutorials, reviews
  • Twitter/X — Text-based influence, thought leadership
  • Pinterest — Visual content, lifestyle and product pins
  • Blogs — WordPress, Substack, Medium, personal sites

Default: Instagram, TikTok, YouTube enabled

Content Type Preferences

Prioritize creators who produce specific content types:

  • Video (Short-form) — Reels, TikToks, Shorts
  • Video (Long-form) — YouTube videos, IGTV
  • Photo/Image — Static posts, carousels
  • Written/Blog — Articles, reviews, tutorials
  • Live Content — Streams, live shopping

Niche Categories

Define which content categories are relevant to your brand:

  • Primary Niches — Must create content in these categories
  • Secondary Niches — Bonus points if they cover these topics
  • Excluded Niches — Never recommend creators in these categories

Categories include: Beauty, Fashion, Fitness, Health & Wellness, Food & Beverage, Tech & Gadgets, Home & Lifestyle, Parenting, Travel, Gaming, Finance, and 50+ more.

Geographic Targeting

  • Creator Location — Where the creator is based
  • Audience Location — Where their followers are located
  • Language — Primary content language

Tip: For best results, align audience location with your shipping regions.

Engagement Thresholds

  • Minimum Engagement Rate — Default: 2% (varies by platform)
  • Comment Quality — AI evaluates comment authenticity
  • Growth Trend — Prefer growing accounts vs. declining

Exclusion Rules

Automatically exclude creators who match certain criteria:

  • Already in Program — Skip existing affiliates
  • Previously Rejected — Don't resurface rejected creators
  • Competitor Affiliates — Exclude known competitor partners
  • Keyword Blocklist — Exclude based on bio or content keywords
  • Brand Safety — Filter controversial or inappropriate content
Brand Safety

Brand safety filters are enabled by default. The agent scans for controversial topics, explicit content, and potential PR risks. You can adjust sensitivity in settings.

Fit Score Algorithm

Every discovered creator receives a fit score from 0-100. The score is calculated using multiple weighted factors:

Score Components

  • Audience Match (30%) — How well their audience matches your customer demographics
  • Content Relevance (25%) — Alignment between their content and your products
  • Engagement Quality (20%) — Authentic engagement, not just volume
  • Brand Affinity (15%) — Prior mentions of your brand or similar brands
  • Historical Performance (10%) — If lookalike to successful affiliates

Score Interpretation

  • 90-100 (Excellent) — Highly likely to convert and perform well
  • 75-89 (Strong) — Good match, worth pursuing
  • 60-74 (Moderate) — Potential fit, review carefully
  • 40-59 (Weak) — Limited alignment, consider if niche is valuable
  • Below 40 — Not recommended, filtered by default

Default threshold: Only creators with 60+ scores appear in recommendations. Adjust this in settings.

Recommendation Queue

Discovered creators appear in your recommendation queue, accessible at AI → Discovery Agent → Recommendations. Each recommendation includes:

Creator Profile Card

  • Profile Summary — Name, handle, bio, profile image
  • Platform Presence — Followers on each platform, growth trend
  • Engagement Metrics — Engagement rate, average likes/comments
  • Fit Score — Overall score with breakdown by factor
  • Content Samples — 3-6 recent posts for quick review
  • Audience Insights — Demographics, location, interests

AI Reasoning

Each recommendation includes an explanation of why the creator was suggested:

  • Key factors that contributed to the fit score
  • Specific content that triggered the match
  • Comparison to your successful affiliates (if applicable)
  • Any potential concerns or considerations

Available Actions

  • Approve — Add to recruitment list, optionally trigger Outreach Agent
  • Reject — Remove from queue, won't be suggested again
  • Save for Later — Keep in separate list for future consideration
  • View Full Profile — Open detailed creator analysis
  • Start Outreach — Immediately begin Outreach Agent sequence
Bulk Processing

Use checkboxes to select multiple creators and apply bulk actions. Great for quickly processing a backlog of recommendations.

Automated Actions

Configure what happens automatically when high-fit creators are discovered. Set these in Discovery Agent → Settings → Automations.

Add to Recruitment List

  • Trigger: Fit score above threshold (configurable, e.g., 85+)
  • Action: Auto-add to a specified recruitment list
  • Use case: Build curated lists for different campaigns

Trigger Outreach Agent

  • Trigger: Fit score above threshold (configurable, e.g., 90+)
  • Action: Automatically start Outreach Agent recruitment sequence
  • Settings: Select sequence template, set daily limits
  • Safeguard: Maximum 25 auto-outreach per day (configurable)

Notify Team

  • Trigger: VIP discovery (e.g., 100K+ followers with 90+ score)
  • Action: Send Slack message or email to team
  • Use case: Ensure high-value prospects get immediate attention

Auto-Invite to Program

  • Trigger: Creator already mentions your brand + high fit score
  • Action: Send program invitation directly
  • Use case: Fast-track existing brand fans into your affiliate program

Lookalike Discovery

One of the Discovery Agent's most powerful features is finding creators similar to your top performers. Here's how it works:

  1. Identify Top Performers — System analyzes your highest-converting affiliates
  2. Extract Patterns — Content style, audience demographics, engagement patterns
  3. Search Database — Find creators matching those patterns who aren't affiliates yet
  4. Score Similarity — Rank by how closely they match your winners
Seed Your Lookalikes

The more successful affiliates you have, the better lookalike discovery works. If you're just starting, manually tag your best affiliates as "lookalike seeds" to jumpstart this feature.

Performance Metrics

Track the Discovery Agent's effectiveness with these metrics, available in the Analytics → AI Performance dashboard:

Key Metrics

  • Creators Discovered — Total new creators found (daily/weekly/monthly)
  • Recommendation Accuracy — % of approved vs. rejected recommendations
  • Approval Rate — % of recommendations you approve
  • Conversion Rate — % of approved creators who become active affiliates
  • Time to Affiliate — Days from discovery to first sale
  • Lookalike Hit Rate — % of lookalike suggestions that match top performer patterns
  • Score Calibration — How well fit scores predict actual performance

Best Practices

Provide Consistent Feedback

The agent learns from every approve/reject decision. For best results:

  • Review recommendations daily during the first 2 weeks
  • When rejecting, select a reason from the dropdown (helps training)
  • Approve creators you'd consider, even if you're not ready to outreach yet
  • Use "Save for Later" sparingly—it doesn't train the model

Tune Your Thresholds

Start with default settings, then adjust based on volume and quality:

  • Too many recommendations? — Raise minimum fit score threshold
  • Too few recommendations? — Lower thresholds or expand platforms
  • Wrong creator types? — Refine niche categories and exclusions

Automate Carefully

Start with manual approvals, then enable automations as you trust the system:

  1. Week 1-2: Review all recommendations manually
  2. Week 3-4: Enable auto-add to list for 90+ scores
  3. Month 2+: Consider auto-outreach for highest scores

Troubleshooting

No Recommendations Appearing

  • Check if filters are too restrictive (especially follower range)
  • Verify niche categories are configured
  • Ensure platforms are enabled
  • Lower minimum fit score threshold temporarily

Wrong Type of Creators

  • Add keywords to exclusion list
  • Tighten niche categories
  • Reject poor matches with detailed feedback
  • Check if brand safety filters need adjustment

Recommendations Seem Stale

  • Agent scans daily; new creators appear continuously
  • Enable more platforms for broader discovery
  • Check if your criteria is too narrow for the database

API Access

Access Discovery Agent programmatically via the Spark! API:

  • GET /ai/discovery/recommendations — Fetch pending recommendations
  • POST /ai/discovery/recommendations/:id/approve — Approve a creator
  • POST /ai/discovery/recommendations/:id/reject — Reject a creator
  • GET /ai/discovery/settings — Get current configuration
  • PUT /ai/discovery/settings — Update configuration

See the API Reference for full documentation.

Next Steps

Was this page helpful?

Need more help? Contact support