How to Find Influencers Manually

What matters most

Write several search angles from the audience's language: category terms, problems, formats, subcultures, locations, games, products, adjacent interests, and recurring events. Platform-native search often reflects the community better than generic web queries.

Follow networks. Strong candidates appear in collaborations, comments, guest appearances, recommended channels, event lineups, community moderators, mutual follows, creator collectives, and the audiences of already-relevant creators.

Capture a canonical profile with handles, URLs, why the creator was found, relevant examples, source date, representative, and contact preference. Discovery evidence should survive beyond a browser tab or an operator's memory.

A practical workflow

1. 01

   Translate the campaign audience and subject into a list of search angles and communities.

2. 02

   Search each relevant platform, the open web, events, publications, and recommendation networks.

3. 03

   Follow collaborations, commenters, guests, mutuals, and competitor sponsorship trails.

4. 04

   Save relevant examples and deduplicate identities across platforms.

5. 05

   Shortlist against fit, evidence quality, availability, safety, and contactability.

Queries that actually narrow the list

Platform-native search returns different results than a web search engine, because it ranks on watch behaviour inside a community rather than on links across the open web. Search the platform first, in the audience's own vocabulary, then use the filters the platform gives you: upload date to surface creators who are currently active rather than historically large, duration to separate long-form reviews from clips, and the caption or feature filters where they exist.

Sponsorship trails are the highest-yield query type, because a creator who has already run a comparable partnership is demonstrably contactable, demonstrably willing, and has a public example of how their audience responded. Search a competitor's brand name alongside the words creators use when disclosing a partnership, and search the category term alongside the format word the community uses for sponsored content.

Web search still earns its place for the surfaces platforms do not index well: newsletter sponsor rosters, podcast guest lists, event and tournament line-ups, community wikis, subreddit sidebars, press coverage of a scene, and creator collectives with their own sites. Restrict the query to a single domain when a publication or community already aggregates the people you want.

Mining the recommendation graph

Once you have three genuinely relevant creators, the fastest route to thirty more is the network around them rather than another search. Every platform exposes some version of it: recommended and featured channels, similar-account suggestions after a follow, collaboration videos, guest appearances, raid and host relationships, tagged co-authors, and the creators who appear repeatedly in each other's comment sections.

Communities are the other half of the graph. Moderators and long-standing members of a Discord server, a subreddit or a forum are often creators themselves, and the people a community reliably links to are a better relevance signal than a follower count. Event line-ups and tournament brackets do the same job for a scene with an offline component: somebody has already assembled a list of the relevant people and published it.

Work the graph deliberately rather than opportunistically. Take your three strongest candidates, exhaust one hop from each, then re-rank before going a second hop out. Two hops from a strong seed is usually still relevant. Four hops is a different market.

The record you keep is the deliverable

Manual discovery produces a list, but the list is worth very little without the evidence that produced it. A record that survives the campaign carries a canonical creator name, every platform handle and URL found, the date and route by which they were discovered, two or three specific pieces of content with links, the language and country signals observed, a sentence on why the audience fits, any safety flags, the contact route and where the creator published it, the representative if there is one, and a status.

Recording the source date matters more than it looks. Contact routes, follower figures and content directions all change, and six months later the difference between something you observed and something you assumed is invisible unless it was written down at the time.

Deduplicate before evaluating, not after. The same person will appear as a YouTube channel, a Twitch handle, an Instagram account and an X profile, and treating those as four candidates inflates the shortlist and can produce four separate outreach messages to one inbox. The reverse error is just as common: a shared handle across platforms is not proof of the same owner, and a channel run by a studio or a team is not an individual creator, which changes both the contact route and the negotiation.

When manual discovery stops paying

Manual work is genuinely the right method up to roughly the first hundred candidates, and it is the only method that teaches you the market's vocabulary, its formats and its unwritten norms. Skipping it means running every later search with the wrong words.

What changes past that point is not finding creators but re-verifying them. Audience composition, posting cadence, contact routes, representation and brand-safety context all drift, and a hand-built list is a snapshot of the week it was made. The cost that eventually justifies tooling is the recurring cost of keeping a list true, not the one-off cost of building it.

Common mistakes

• Searching only a broad hashtag and accepting the largest accounts.
• Treating search rank as proof of audience or brand fit.
• Creating separate records for every handle owned by the same creator.
• Saving names without the content evidence that made them relevant.

Working checklist

• Search terms reflect the audience's language and adjacent interests.
• Multiple discovery channels were used.
• Each candidate has a reason and supporting content example.
• Cross-platform identities are deduplicated.
• The shortlist is ready for consistent evaluation.

Questions and answers

How many creators should a shortlist contain?

Work backwards from the reply rate rather than picking a round number. Cold outreach to creators who have a published business contact and a genuine fit loses candidates at three stages: no reply, a decline or a rate mismatch, and a fit or safety issue found during closer review. Building a shortlist several times the size of the roster you intend to sign absorbs all three without a second discovery pass.

How do you find creators who have not worked with brands before?

Look for the absence of the signals sponsored creators accumulate: no media kit, no rate card, no business email in the profile, no disclosure language in recent uploads, no representative listed. Community spaces and recommendation graphs surface them more reliably than search does, because they rank on audience behaviour rather than on the discoverability work a commercial creator has already done.

Is it acceptable to build a list from a competitor's sponsored content?

Yes. Disclosed partnerships are public information, and reviewing who a competitor worked with is ordinary market research. Two cautions: a creator may be under an exclusivity or category-restriction clause that makes them unavailable for a defined period, so ask rather than assume, and a list built only from a competitor's roster inherits that competitor's audience assumptions along with their creators.

How do you tell a real creator from a repost account?

Look for evidence of a person doing work: on-camera or on-mic presence across uploads, a consistent voice and setting, replies in their own comments, content that references its own history, and content that could not have been reposted because it is about the creator's own life or process. Compilation and repost accounts can post real numbers and are simply a different product, so decide deliberately rather than by accident.