6 min read
Three ways to get your League of Legends match history into ChatGPT — an op.gg link, a screenshot, or text/JSON — compared. Why link-pasting quietly fails, the copy-paste steps that work, and the first prompt to send with your data.
This article is currently shown in English. A translation is in progress.

CHAMPION LENS
Ezreal
Lux
Yasuo
Blitzcrank
Morgana"Show ChatGPT my match history and have it tell me what's wrong" — the goal is simple, but the moment you go to paste something, you hesitate.
Do you paste your op.gg link? A screenshot of the post-game screen? Type the numbers by hand?
That choice largely decides the quality of the analysis you get back. This guide compares the three methods and lays out the reliable one as copy-paste steps.
| Method | Effort | Accuracy | Verdict |
|---|---|---|---|
| Paste an op.gg link | Easiest | Poor | Not recommended — often unreadable, numbers missing |
| Paste a screenshot | Easy | Shaky | Supplement only — misreads and missing context |
| Paste text (JSON) | Easy | Exact | Recommended — numbers arrive intact, analysis gets specific |
The most common attempt is pasting your op.gg profile URL with "analyze this."
ChatGPT can't always open links, and even when it can, stat-tracker pages render their numbers with scripts after the page loads — so the actual stats often never reach the AI.
The worst part: even with no data, the AI will still produce a plausible-sounding analysis. It's easy to mistake an invented impression for a real read of your games.
If you do paste a link, always follow up with "what were my KDA and CS?" If it can't answer, it never read the page.
A post-game screenshot is easy and far better than a link, but it has two weaknesses.
Misread numbers: image recognition can read CS 187 as 167, and every comparison downstream inherits the error.
Missing context: a screenshot only shows final stats — no death timings, no timeline against your lane opponent.
Screenshots work fine as extra material — showing a build, for instance — but shouldn't be the primary data for the analysis.
The reliable route is pasting your stats as text. The numbers arrive as exact characters, so nothing gets misread or guessed.
You could transcribe by hand, but LoL2LLM gets you paste-ready text in four steps:
Search your Riot ID (GameName#Tag) on the home page
Pick the match you want analyzed from your recent games
In the export settings, check the fields you want to pass (KDA, CS/min, damage share, vision, …)
Hit copy and paste the JSON into ChatGPT
No account needed, and the output is trimmed to a size the free ChatGPT tier handles comfortably. What each JSON field means is covered in the JSON fields guide.
Paste the JSON alone and the AI tends to summarize it back at you. Always attach an ask before or after the data:
Below is my League of Legends match data (JSON). Identify my role and lane opponent, name the ONE metric where the gap was largest this game, and suggest 2 improvement actions I can try next game. Mark anything not present in the data as speculation. [Paste the copied JSON here]
The "mark anything not in the data as speculation" line is the cheapest insurance against hallucination.
For going deeper — the role / ask / constraints prompt pattern — see how to analyze your games with ChatGPT, and for role-specific prompts, the prompt collection.
Because the export is trimmed to the fields that matter, your last five games together still fit ChatGPT's and Claude's context comfortably.
Multiple games actually beat a single one — the AI can extract the weakness that repeats across games instead of reacting to one lucky or unlucky match.
Start by exporting one game with your own Riot ID and sending the prompt above as-is.
Not recommended. ChatGPT can't always open links, and stat-tracker pages render their numbers with scripts, so the actual stats often never reach the AI. Worse, it will still produce a plausible-sounding analysis with no data behind it. Paste your stats as text (JSON) instead.
Text (JSON). Screenshots suffer from image-recognition misreads and only show final stats, losing death timings and the timeline. They're fine as a supplement for things like builds, but the primary data should be text.
Yes. The LoL2LLM export is trimmed to the fields needed for analysis, so one to five games fit comfortably within the free tier's context. No account registration is needed either.
Yes. Pasting the JSON with an attached ask works identically across ChatGPT, Claude, and Gemini. Whichever AI you use, adding 'mark anything not present in the data as speculation' keeps hallucination down.
AhriA step-by-step method for getting coach-level analysis of your League of Legends games from ChatGPT or Claude. Why screenshots fail, how to feed structured data, and role-by-role prompts you can copy.
(English)
OriannaA field-by-field walkthrough of the JSON LoL2LLM hands to ChatGPT or Claude. Knowing what you're sending makes your prompts — and the AI's answers — much sharper.
(English)
JhinWhen to use which of LoL2LLM's five personas, plus the one-line additions that turn a generic answer into a sharp one.
(English)