Video Diagnostic

Why isn't this video getting views?

Paste a video, add your analytics, and get a diagnosis of exactly where viewers drop off — and what to fix.

Video

Add data

Diagnosis

Paste the link to the video you're worried about. We pull its title, thumbnail and transcript — then you add your Studio analytics for the full diagnosis.

YouTube video link

"It flopped" is not a diagnosis

When a video underperforms, the useful question is not whether it failed but where. A video that got no impressions failed at discovery. One that got impressions and no clicks failed at packaging. One that got clicks and lost everyone at forty seconds failed at the hook. Those are three unrelated problems with three unrelated fixes, and the generic advice to "improve your titles and thumbnails" will help with at most one of them.

This tool maps a single video against six funnel stages — packaging, discovery, click-through, hook, retention, and engagement — scores each one, and then does the thing that actually matters: names the single stage that broke first.

Three passes, then a verdict

Public data is fetched for free before anything is charged

Paste a URL and the tool retrieves the video's title, description, tags, duration, publish date, engagement counts, and channel subscriber count from YouTube's official Data API, plus the transcript where captions exist. This step costs nothing — you can see what was collected before committing to an analysis.

Your analytics screenshots are read by a vision model

You can attach several screenshots from YouTube Studio. A vision pass extracts click-through rate, impressions, unique viewers, average view duration and percentage, watch time, subscribers gained and lost, traffic sources, and the retention curve. The instruction governing this pass is strict and worth quoting in spirit: extract every metric you can see, and where a value is not visible, return nothing rather than guessing. Every numeric field is allowed to be empty, which is what stops the diagnosis inventing a CTR it never saw.

A reasoning pass applies explicit thresholds and picks one bottleneck

Everything collected is handed to a reasoning model working against stated numeric thresholds — the rubric treats a mid-single-digit click-through rate as broadly normal and something below about three percent on meaningful impressions as a click problem, and flags an average view percentage under roughly thirty percent as weak for most formats. Its defining instruction is to identify the single earliest stage that breaks and explicitly not to spread blame evenly, because fixing a later stage is wasted effort when an earlier one is the wall.

What it cannot see without your help

The single most important thing to understand about this tool is where its data comes from — and does not.

It has no connection to your YouTube Analytics account. There is no login, no OAuth, and no data feed. Click-through rate, impressions, and retention exist in the diagnosis only if you upload screenshots, and they are read off pixels by a vision model.
Without screenshots the diagnosis still runs, but it works from public signals only and is required to state plainly that CTR and retention could not be measured. That is a genuinely weaker diagnosis, and the tool says so rather than bluffing.
Screenshots must be legible. A blurry or heavily cropped capture yields empty fields, and empty fields lower the confidence of the stages that depend on them.
There is a limit on how many screenshots one run accepts, so upload the views that carry the most information — the overview panel and the retention curve first.
Long transcripts are truncated, front-loaded toward the opening because that is where hook analysis needs the detail. Videos without captions get no transcript at all, and the hook is then judged from title and description alone.
Scores are model judgements against the thresholds above, not benchmarks against a measured dataset of comparable videos.
It diagnoses one video. For channel-level patterns across many uploads, use the Channel Audit instead.

Diagnosing a video properly

1
Paste the video URL
Drop the link to the video you're worried about. We pull its title, thumbnail, and transcript automatically.
2
Add your Studio analytics
Upload screenshots of the video's YouTube Studio analytics — CTR, impressions, retention, and traffic sources — so the AI can see where the funnel breaks.
3
Read the diagnosis
Get an overall health score, the single primary bottleneck, a stage-by-stage scorecard, prioritized fixes, and quick wins — exportable as a PDF.

What to do with the verdict

The output is designed to send you somewhere specific. Roughly, by bottleneck:

Discovery — the topic or metadata is the problem. Go to keyword research and then re-package with the SEO Optimizer.
Packaging or click — impressions arrived and nobody clicked. Score the image in the Thumbnail Analyzer and rebuild it in the Thumbnail Generator.
Hook — they clicked and left immediately. Take the script to the Hook Optimizer.
Retention — they stayed a while and drifted. That is a pacing and structure problem, and it is fixed in the next script rather than in this video.
Engagement — people watched and did nothing. Look at what you asked for, and check what the audience is actually asking you for in the Audience Voice analyzer.

Video diagnosis questions

How is this different from the Channel Audit?

Channel Audit scores your whole channel. Diagnose drills into one specific video — reading its title, thumbnail, transcript, and your Studio analytics to find the exact funnel stage that's failing.

Do I need to upload analytics screenshots?

No, but they make the diagnosis far more accurate. Without them the AI works from public signals and clearly flags that CTR and retention couldn't be measured.

What does the diagnosis cover?

Six funnel stages: packaging (title × thumbnail), discovery & SEO, click-through (CTR), opening hook, retention & pacing, and engagement — plus the single biggest bottleneck and how to fix it.