Find and Vet Influencers with Claude or ChatGPT

Can Claude or ChatGPT check an influencer before a sponsorship?#

Yes, if the assistant can read the influencer's account. Connected to live public data, Claude or ChatGPT can check four things in a few minutes: whether engagement is steady or propped up by a few posts, whether comments come from real people, what a sample of followers looks like, and how the creator performs on Instagram versus TikTok. It cannot see the creator's reach or audience demographics; for those, ask the creator for screenshots of their Insights.

Updated October 2026.

This guide shows how to find candidates and how to vet them with AI, using StoriesFly as the connector. For the wider process (briefs, contracts, rates, measuring results), see the Instagram influencer marketing guide.

Setup#

Add StoriesFly to your assistant once. Steps for each app are on the connect page.

• Claude: Customize → Connectors → Add custom connector → https://storiesfly.com/mcp → Connect → sign in with Telegram.
• ChatGPT: Plugins → Add → Add custom MCP server → same address → OAuth → Connect → sign in with Telegram. Works on the Free plan.

The first 100 requests are free, once per Telegram account: enough to vet three or four creators properly. You never log in to Instagram or TikTok with your own account, and your account never appears among a story's viewers. For more details, see our guide on view stories anonymously.

Part 1: Finding candidates#

AI is useful for building a long list, but it can only search what the tools expose. Three ways that work:

Search by name or niche word (1 request per page):

Search Instagram accounts for "vegan baking". Show the first page
with follower counts, and drop anything over 200K followers.

Hashtag top posts (about 2 requests per hashtag):

Read one page of top posts under #veganbaking and #plantbaseddessert.
List the accounts behind them with follower counts.

Who your competitors already work with (1 request per page):

Read the first page of accounts @competitorbrand follows on Instagram,
and the latest TikTok reposts of @competitorbrand. Which look like creators?

On TikTok, reposts are a good signal: a brand or creator reposting someone usually means they work together or like the work. StoriesFly does not search TikTok by keyword or hashtag, so start TikTok lists from accounts you already know.

Part 2: Vetting one creator#

Here is a full check for one creator, with request counts. Account name made up.

Engagement stability#

Read the profile and the last 36 posts of @plantbaker (4 pages).
Compute engagement rate per post (likes + comments / followers).
Show the median, and how many posts fall below half the median.
Is the median held up by a few viral posts?

Cost: 5 requests.

What you are looking for: a steady band of engagement. A creator with a 4% median where most posts sit between 3% and 5% is reliable. A creator with a 4% average driven by two viral posts and the rest at 0.8% will probably deliver 0.8% on your post.

Also ask about the trend:

Split the 36 posts into the oldest 18 and the newest 18.
Did engagement go up or down?

Comment quality#

Read the comments on her 3 most recent posts (1 page each).
What share are generic ("Nice!", emoji only, "Love this ❤️"),
what share mention the actual content, and are any accounts
commenting on every post with similar text?

Cost: 3 requests.

Real audiences ask questions, tag friends and argue a bit. A wall of short generic praise, often from the same handful of accounts, points to engagement groups or bought comments.

Follower sample#

Read the first 100 followers of @plantbaker (4 pages).
How many have no profile picture, no posts, a name made of random
letters and numbers, or follow thousands of accounts?

Cost: 4 requests (a page holds up to 25 accounts).

This is a sample, not an audit: 100 accounts out of many thousands. If a large share look empty, ask the creator about it. Note that for big verified accounts Instagram shows only about 50 followers.

Likers sample#

Read who liked her latest post. Same checks as the followers.

Cost: 1 request. Instagram shows only a sample of likers on busy posts, so this is a spot check.

TikTok as well#

@plantbaker is on TikTok with the same name. Read her profile and
latest videos. Median views per video vs her Instagram reels?
Any outlier videos (more than 3× her median)?

Cost: about 3 requests.

A creator strong on one platform and weak on the other is common. It tells you where to put the budget.

The verdict#

Summarize: engagement stability, comment quality, follower sample,
TikTok vs Instagram. Then give me 3 questions to ask her before we sign.

Cost: 0.

Total for one creator#

Vetting a shortlist of ten creators is about 160 requests. That fits Pro (500 a month for $19.99) or a prepaid balance at $1.50 per 1,000 requests at the entry price, with no monthly cap and a balance that never expires. Basic gives 100 a month for $4.99. See plans.

What public data cannot show#

Be clear with your team about this before anyone signs off:

• Reach and impressions. How many people actually saw the posts. Only in the creator's Insights.
• Audience demographics. Country, age and gender of followers. Only in Insights. If your product ships to one country, this matters more than anything above.
• Story views. Not public.
• Saves and shares on every post. Partly available per post on Instagram where Instagram gives them; not as a full history.
• Past sponsored results. Clicks, codes used, sales. Ask for case studies.
• Private accounts. Not readable. A creator you want to pay should have a public account anyway.

The standard practice is to ask the creator for screenshots of Insights for the last 30 days (reach, audience countries, age split) and compare them with what the public numbers suggest. If a creator with 100K followers shows reach of 5K per post, the follower count is not the number to pay for.

Red flags the assistant can catch#

• A large follower count with very few likes and comments on recent posts.
• Engagement median far below the average: a few viral posts carry the account.
• Comments mostly generic, or the same accounts on every post.
• A large share of the follower sample with no posts and no picture.
• Long gaps in posting, then bursts.
• Big numbers on one platform, very low on the other with the same content.

None of these proves anything alone. Two or three together are a reason to ask questions or move on.