analyze_trend scores a public post; it is not prompt-match QC
Viral-post analysis is one named research job at 1 credit. Using it to judge your generate, or using review_generation to explain someone else's Reel, answers the wrong question.
analyze_trend scores a public post. You pass a URL. You get a viral score, hook analysis, content breakdown, suggested caption and hashtags, recreation tips. It costs 1 credit per analysis. It is not a check of whether your generate matched your prompt. It is not "what's trending in fitness." It is not a virality predictor on an unpublished file. Share a generation URL and ask why it will go viral and you have spent the research credit on a question the tool was not built to answer.
Analyze a viral video or post is the named agent job. Share the link. Get the breakdown. Required param: url. That is the entire contract.
What the job is
Someone else's (or your already-public) Instagram / TikTok / similar link. Why it is performing. What the hook is doing. How you might recreate the format, not clone the pixels. Recreation tips are advice. They are not a workflow clone and they are not a hook pack.
If you actually want to clone a public Versely template, recreate a viral template is that job. If you want a pack of openings for your brand, that is create_hook_pack. Analysis does not generate those.
One credit is small, which is why people spam it. The waste is rarely the 1. The waste is the generate you run next from a misread, or the QC you skipped because you thought this was QC.
What it is not
- Prompt-match on a generate. Get AI feedback on a generation is
review_generation: matches / issues / severity / suggested fix against the original prompt. Different input (your media + your prompt). Different question. - A deep breakdown of a file you have. Get a detailed breakdown of a video is
analyze_video: frames, beats, on-screen text, optional transcript. No public URL required. No viral score. - A trend search. Find what's trending is
get_trending_content: platform plus keyword or hashtag, a list of posts with stats.analyze_trendneeds one URL you already have. - Your account analytics. Check my social media performance reads your posts. A viral stranger's Reel is not your dashboard.
Reaching for the wrong row gets you a confident answer to a question you did not ask. Then you spend a model credit "fixing" something that was never the failure.
The test
Do you have a public post URL, and is the question "why is this working / how would I recreate the format?"
Yes: this job. One credit. Read the hook analysis. Recreate on a generate or pack job after, on purpose.
If the file is yours and unpublished, review or analyze_video. If you do not have a URL yet, search trending, then analyze a specific result. Do not ask analyze_trend to invent the URL.
FAQ
Can I pass a Versely generation URL instead of TikTok?
The tool expects a social post URL. A private generate is not a performing post. Use review_generation for prompt match, analyze_video for a beat sheet.
Is the viral score a publish/don't-publish gate for my own cut?
No. It is a reading of a public link. Comparing two of your unpublished cuts is a different product (a predictor on a file). Do not treat this 1-credit research call as that gate.
Does analysis clone the video into my account?
No. Tips are text. Cloning a Versely public workflow is recreate_public_template. Generating your own version is whatever create job the format actually needs.
Why not run this on every post I like?
You can, at 1 credit each. The diminishing return is real: five breakdowns of the same hook structure will not make the sixth generate better. Pick one pattern, then leave research for a named create job.