Enforcement is automated
Platform systems match pixels and audio. They don’t read your licence email, your intent or your fair-dealing argument.
Now onboarding design-partner agencies
Caegis reviews your creators’ YouTube and Instagram videos before they go live. Every finding comes with the timestamp, the official rule it relates to and a suggested fix.
Illustrative example report
Morning routine with FitFuel
Instagram Reel · 0:58 · English
Paid agreement with FitFuel: no disclosure in the opening caption
Fix: start the caption with “Ad:” and turn on the paid-partnership label.
Broadcaster logo on screen: “Star Sports”
“Guaranteed to burn fat in 7 days”
Footage also appears under another client
2.5B
Content ID copyright claims on YouTube in 2025 — almost all of them automated.
Source: TorrentFreak / YouTube transparency report3 strikes
in 90 days and a YouTube channel is subject to termination, with every video on it.
Source: YouTube HelpObviously identifiable
The ASA/CAP Code requires advertising, including influencer posts, to be obviously identifiable as advertising.
Source: ASA, recognising ads in social mediaCMA guidance
The CMA can take enforcement action where advertising is not obviously identifiable as advertising, including hidden endorsements.
Source: CMA, social media endorsements guidanceIndustry figures from the linked public sources, not Caegis customer data.
The problem
An account manager checking 40 videos a week can’t read every caption against platform policy, spot a sponsor named on screen without a label, or remember that the same B-roll went out on another client’s channel last Tuesday.
Platform systems match pixels and audio. They don’t read your licence email, your intent or your fair-dealing argument.
In 2025 YouTube and Meta both stepped up action against reused, unoriginal content.
In the UK, the ASA can rule against hidden advertising under the CAP Code, and the CMA can take enforcement action where an ad is not obviously identifiable.
Deleting a video after the fact doesn’t remove the strike. Problems are cheapest to fix before the post goes live.
How it works
Upload the final cut (up to 30 minutes) with its caption and hashtags, or import a post from a creator account you’ve connected through the official Instagram or YouTube APIs. No scraping.
Speech is transcribed, key frames are sampled scene by scene, and on-screen text, handles, watermarks and logos are read.
Deterministic checks plus an AI review that can only cite a versioned rule library, written from official YouTube, Instagram and Meta pages and being extended with the ASA/CAP Code and CMA guidance. Uncited findings are discarded.
Get timestamped findings with evidence, the policy citation and a suggested fix, plus a shareable HTML/JSON report and an evidence pack if you ever need to dispute or appeal.
Features
Everything below is in the product today. Where something is optional, paid or on the roadmap, we say so.
Community guidelines, advertiser-friendly and monetisation rules, spam and misleading metadata, medical misinformation, AI-content labels — for YouTube and Instagram.
Finds missing or buried “Ad”, “Advertisement”, “Gifted” and #ad labels, sponsors named on screen or in speech, and disclosures hidden among hashtags — aligned to the ASA/CAP Code and CMA guidance.
Each finding is pinned to the moment it happens, quotes the evidence (speech, caption or on-screen text) in the original script, and suggests a concrete edit.
Video and audio fingerprints flag footage that already exists in your catalogue — including mirrored copies and the same clip under two different clients.
Record licences, permissions and brand agreements with proof files. Findings reference the ledger, so a licensed clip or a disclosed deal shows up as covered.
One zip with the report, the file’s SHA-256, every match, the ledger records it relies on with their proof, and a draft dispute note.
Flags policy, disclosure and reuse issues that can put a brand campaign at risk, so account teams can fix them before the post goes live.
Uploaded media is deleted after 72 hours by default; only fingerprints are kept for matching. Connected-account tokens are encrypted. Each agency runs on its own dedicated instance.
Music identification and web image search for stock, broadcast or film stills can be switched on when a signal calls for them. Off by default, with a monthly spend cap.
For UK agencies
UK agencies publish sponsored YouTube and Instagram content under the ASA/CAP Code and CMA guidance: advertising must be obviously identifiable as advertising. A missing or buried ad label, or the same clip used for two clients, is a problem for the brand as well as the creator. Caegis checks the video before it is published and cites the rule it relates to.
Platform policy and ASCI disclosure rules for Indian agencies.
AvailableYouTube & Instagram policy checks, repost matching, rights ledger and evidence packs work today. FTC Endorsement Guides coverage is rolling out.
FTC rules: rolling outYouTube & Instagram policy checks, ASA/CAP Code and CMA-aligned disclosure, repost matching, rights ledger and evidence packs. ASA/CAP coverage is rolling out with UK design partners.
Design partners welcomeCaegis vs. manual review
| What gets checked | Manual review | With Caegis |
|---|---|---|
| Every second of speech | Spot-checked, depends on the reviewer | Transcribed in English, each issue timestamped |
| On-screen text, handles, logos | Easy to miss in fast edits | Read from scene-aware key frames |
| Sponsorship disclosure | Checked against memory of the rules | Checked against the ASA/CAP Code, CMA guidance and platform rules, cited |
| Same footage used for another client | Practically impossible to track | Fingerprint match across your whole catalogue |
| Licences & permissions | Scattered across email and drives | Rights ledger, linked to findings |
| Proof when a claim or strike lands | Assembled after the fact | Evidence pack generated per review |
| Judgement calls (satire, context, intent) | Your team | Still your team — Caegis flags, people decide |
What it catches
These are made-up examples to show the kind of output you get — not customer stories.
Fitness creator, paid shake deal
Risk: #ad is the 14th hashtag at the end of the caption. The CAP Code requires advertising to be obviously identifiable — a disclosure buried among hashtags may not count.
Caegis: Flagged high — “Paid agreement with FitFuel: opening disclosure needs review.” Fix: start the caption with “Ad:”.
Agency managing 12 creators
Risk: The same stock transition clip goes out on two clients’ channels in one week — one copy unlicensed.
Caegis: “Footage also appears under Client B”, with both timestamps, before the second post.
Cricket edits page
Risk: Highlights with a broadcaster’s scorebug visible from 0:12 to 0:41.
Caegis: Flagged as third-party broadcast footage. Record the licence in the ledger, or cut the segment.
No tool can make a channel “ban-proof”. Platforms decide, their automated systems make mistakes, and rights-holders can claim content at any time. Caegis catches the problems it can see before you post, shows the evidence and cites the official rule, so your team decides with its eyes open. The risk indicator is a heuristic, not a probability, and nothing here is legal advice.
FAQ
No, and nobody honestly can. We catch what is visible in the video, caption and metadata before posting, and give you evidence and citations. Final decisions sit with the platforms — and with your team.
No — use both. YouTube’s checks look for Content ID matches and ad suitability after you upload, on YouTube only. Caegis runs before upload, covers Instagram as well as YouTube, and also checks community guidelines, sponsorship disclosure, claims and your own licences.
From a versioned library of 35 rules, each written from an official YouTube, Instagram or Meta page that was read and dated, and being extended with the ASA/CAP Code and CMA guidance. The AI can only cite rules in that library; anything it can’t tie to a rule is thrown away.
No. You upload a file, or connect your own accounts through the official Instagram and YouTube APIs. YouTube’s API terms don’t allow downloading videos, so for YouTube you attach your own copy and we cross-check it against the title, description and captions.
Uploaded media is deleted automatically after 72 hours by default, and you can delete it sooner. Large files are converted to a working copy with location metadata stripped. Only fingerprints are kept for repost matching, and you control how long.
Yes. Disclosure checks are being aligned to the ASA/CAP Code and CMA guidance, alongside YouTube and Instagram policy risk, reused footage, the rights ledger and evidence packs. Fuller ASA/CAP coverage is rolling out with early UK design partners.
We set up a dedicated Caegis instance for your agency, onboard a handful of your creators, and run real pre-publish reviews for 30 days. You tell us what was useful and what was noise. See pricing for pilot terms.
Today each agency instance has one shared operator login. Individual reviewer accounts and roles are on the roadmap; tell us how your team reviews and we’ll factor it in.
We’re taking a small number of design-partner agencies in the UK, and we also work with agencies in India and the US.