The links a creator publishes in a public bio span several different categories. A typical bio holds between four and ten URLs covering their other social accounts, a Linktree or bio aggregator, sponsor or affiliate links, and sometimes their personal website. The personal site is the one a brand cares about: it is the creator's own space, where they speak in their own voice and where partnership inquiries are received. Most of those URLs are noise for this purpose; the steps below identify the one that belongs to the creator.
Creators publish URLs in several public-facing places. Bio text is the most visible: creators paste links directly into the bio for anyone to see. Other publicly visible surfaces on each platform are considered alongside the bio.
Most URLs in a bio aren't personal websites. They're other accounts, link aggregators, marketplace listings, or sponsor links. The pipeline maintains curated deny-lists for each category and removes any URL whose domain falls into one of them.
Links pointing to the creator's other social profiles. Useful elsewhere on the platform but not their personal site.
instagram.com · youtube.com · tiktok.com · x.com · snapchat.com · …Hosted bio pages that just redirect to other links. Not the creator's owned site.
linktr.ee · bio.fm · beacons.ai · stan.store · …Sponsor links, affiliate URLs, marketplace listings. Often shortened or tracking redirects.
amzn.to · etsy · hotmart · calendly · …A small set of URLs that don't match any deny-list category. These are the candidates for the next step.
nikolatravels.com · borislifelog.co · …A URL that survives the filter still needs to be linked to the creator. The domain and subdomain of each candidate are checked against the creator's known identity: their username, full name, and email local part. The match uses both exact substring checks and fuzzy similarity, so partial matches and slight variations still resolve.
A creator named "NikolaTravels" might have nikolatravels.com, or nikola-travels.com, or travelswithnikola.co. The exact substring won't always match the handle character-for-character. Fuzzy similarity catches the variations while the deny-lists from Step 2 already eliminate the obvious non-matches, so we can be permissive at this stage without compromising the result.
Identity match alone is not enough. A lookalike domain or a fan-built tribute site can pass the substring or fuzzy check coincidentally. Each candidate site that survives Step 3 is verified with AI to confirm it belongs to the creator. The AI classifies the site (creator-owned, sponsor or business landing page, fan site, parked domain), and only sites confirmed as the creator's own move forward.
Personal voice, the creator's own content, partnership inquiry forms, original work. The site is confirmed and moves to the ranking step.
Sponsor pages, generic business landing pages, fan-built tribute sites, parked domains. These are filtered out before ranking.
When AI review surfaces meaningful information from the site (additional contact addresses, a media kit, evidence the creator sells products, monetization signals), it is used to update the creator's other data points. A confirmed personal site can feed Email, Has Merch, Income, and other contact-related signals without requiring a separate refresh.
A creator can have more than one personal site, for example a portfolio alongside a newsletter, or a video hub alongside a merchandise storefront they own directly. Every URL that clears the filter and the identity match is retained. The strongest match (typically the one whose domain corresponds most closely to the creator's primary handle) is designated as the primary, which is what the platform uses for outreach, enrichment, and discovery. The full list stays accessible.
A confirmed personal website isn't isolated. It feeds Profile Matching as a signal in three ways.
When nikolatravels.com appears as a confirmed personal site on all three of Nikola's platforms, the shared domain is strong evidence the three profiles belong to the same creator. Profile Matching weighs this alongside link-in-bio agreement and the other signals it uses.
Once Profile Matching has linked a creator's accounts together, the personal sites detected on every platform are merged into one ranked list. Sites surfacing on multiple platforms get a corroboration boost.
The Step 3 match against username, full name, and email local part is what separates a creator's personal site from a random URL in their bio. That same identity check is part of how Profile Matching determines whether two profiles belong to the same creator.
Enrichment: found additional email and social links on the site that feed back into the creator's profile.
The Linktree links can still feed into profile matching, but the personal website field stays empty.
Link-in-bio services (Linktree, Beacons, Stan Store, and similar) are intentionally excluded from the personal website field because they are aggregators, not owned domains. They host the creator's links but the page itself is not the creator's property or creative output. Creators who use these still benefit from the links being read for profile matching (the URLs inside the Linktree page can surface cross-platform handles, emails, and other signals), but the personal website field stays empty until an actual owned domain is found.
Personal website detection is validated using the same approach as every other data point on this site. Two independent tracks run in parallel, AI for scale and human reviewers for nuance, and the deny-lists from Step 2 are reviewed and expanded continuously as new aggregator services and affiliate networks appear.