Find the creators who
feel like your brand.

• The map now opens creator identities, source links and supported native players instead of title-only tooltips.
• The old “stated brand” centroid has been replaced by a clearly labelled website-informed proposal. Advertising is a comparison, not a definition of the ideal brand.
• Inferred items and a disputed source are excluded from the default view. Search pages and campaign coverage are distinguished from individual posts.
• Claims that creator drift causes weaker performance, or that repeating fewer creators solves it, have been removed. This study does not test either proposition.
• Public websites can suggest a direction. Consumer, competitive and commercial evidence are still needed to validate the right space.

1. Confirm market, product, audience, campaign objective and three or four target feelings with the brand owner.
2. Collect a consistent recent sample per creator: for example, 10 posts in 90 days, including organic and sponsored work. Record all selection rules and unavailable posts.
3. Review full video, audio and caption. Score creative evidence with the same rubric, recording source, date, reviewer, confidence and reasons.
4. Check audience geography and relevance, recent distribution, prior brand work, availability and usage rights. A public profile is not enough to establish these.
5. Separate emotional fit, creative consistency, audience relevance and evidence coverage. Do not hide unknowns inside a precise-looking composite score.
6. Test a small paid pilot. Measure brand effects and business outcomes independently before expanding the relationship.

The historical model assigned words from descriptions, using three passes and retaining words supported by two passes. The source script truncates creator descriptions to 420 characters. The placement model has not been validated for this category and task against a consumer panel. Repeated reads do not correct missing stimulus evidence.

This is a convenience sample spanning different dates, formats and objectives. It includes repeated creators, campaign summaries and discovery pages. The audit also found two duplicated TikTok post IDs under different URL variants; 209 is an item count, not a unique-post count. It cannot establish population effects, causal impact on sales, an optimal roster size, or how every creator’s current content feels. A creator should never be characterized personally by a post’s coded attributes.

Website review: 21 September 2026, US pages. The review covered accessible page text, headings and image descriptions, not every website video. Original creator metadata is historical and needs rechecking.

Sigil subsequently connected and opened original TikTok and Instagram posts. The library now includes 29 distinct reviewed previews: 18 mapped items and 11 additional candidate examples. After the initial Sigil pass, 12 more images and captions were checked directly in the native browser. Expanded candidate profiles let you switch between examples and read their individual briefing implications. Cyan rings identify posts with a reviewed image and caption. These checks do not replace full-video and audio review, and the historical map coordinates have not been rescored. Six YouTube references were checked using available transcripts and sampled visual frames. This is not a continuous full-video review or a complete recent-content audit of any creator. Embedded posts may be restricted, removed or require platform login. Every player has a direct-source fallback.

Creator discovery guidance uses Modash documentation, TikTok One and Instagram Creator Marketplace. Players follow TikTok and YouTube documentation.