Steam reviews are some of the most honest signals you can get in PC gaming. A thumbs-up or thumbs-down from someone who paid for the game, played it, and bothered to write something — that's not a marketing-shaped focus group. The problem is the storefront. You see fifteen reviews at a time, sorted however Valve felt like sorting them, and the moment you try to study a review-bomb spike or a localization gap you're clicking forever. If you actually want to look at the data, you need it in a spreadsheet. This guide walks through using ExportComments' Steam Reviews exporter to pull every public review for a title into Excel, CSV, or JSON in one go.

Why export Steam reviews

Steam doesn't do stars. Every review is binary — Recommended or Not Recommended — and the big number on the store page is just the percentage of thumbs-ups. That actually makes the raw data more interesting than a 4.2/5 average, because all the signal lives in the text, the playtime context, and the votes other players left on each review.

Once the dataset is sitting offline, you can do things the storefront simply won't let you do:

  • Detect review bombing by plotting daily volume against your patch dates and external events. The Helldivers 2 PSN-account spike in 2024 is the textbook example — a clean before/after you can chart in five minutes if you have the file.
  • Filter to reviews with more than two hours of playtime to skip the refund window and the drive-by votes.
  • Bucket by language to see which markets are souring on you. A game like Cyberpunk 2077 at launch had radically different sentiment in different locales — the file makes that visible.
  • Compare playtime at review against current playtime to find players who left a thumbs-down and then kept playing anyway. That cohort is interesting.
  • Slice the dataset into pre-patch and post-patch windows to measure whether a major update actually moved the needle on sentiment.
  • Sort negative reviews by helpful votes. The most-upvoted criticism is, in effect, the feedback your community has co-signed — and it's almost always the most actionable thing your design and QA teams can read.

How to export Steam reviews — step by step

Step 1: Grab the Steam store URL

Open the storefront page on Steam. For example: https://store.steampowered.com/app/1086940/Baldurs_Gate_3/. Any canonical app URL works — you don't need to click into the reviews tab first. The numeric AppID is the only piece the exporter cares about.

Step 2: Paste the URL into the exporter

Open the Steam Reviews exporter and drop the URL in. Got several titles to pull at once — say a publisher's full back catalog? Switch to bulk mode and paste one URL per line. Bulk runs return one file per URL, bundled in a single ZIP at the end of the job, so each game stays in its own clean sheet and you don't have to split anything apart later.

Step 3: Pick a format and language scope

Excel (.xlsx), CSV, or JSON. Pick Excel if you want to pivot, filter, and chart immediately. Pick JSON if you're piping the export into a notebook or a sentiment model — it's faster to load and the strings come out clean. By default the exporter pulls every language Steam exposes for that title. Leave it that way unless you specifically need just your home market — once the file is local you can filter on the language column anyway.

Step 4: Start the export

Hit Export. The job runs server-side and paginates through Steam's review feed until it's collected every public review. A title with hundreds of thousands of reviews takes a few minutes — close the tab and walk away. The file lands in your dashboard and your inbox when it's done.

Step 5: Open the file

Open the .xlsx in Excel, Numbers, or Google Sheets. One row per review, and the columns are documented in the next section.

Inside the export — what fields you get

Each row is a single review. You'll find columns for:

  • Review ID — Steam's internal recommendation ID.
  • Author name and SteamID64 — the persistent 17-digit identifier that survives display-name changes.
  • Profile URL — direct link back to the reviewer.
  • Recommendation — Recommended or Not Recommended (the thumbs-up / thumbs-down).
  • Review text — the full body of the review.
  • Language — the language code Steam tagged the review with.
  • Playtime at review (minutes) — a snapshot of how long the author had played when they posted.
  • Playtime total (minutes) — current lifetime playtime for the same author on this title.
  • Helpful votes and Funny votes — the two community-vote counters.
  • Comments count — how many replies the review attracted.
  • Steam purchase — true if the copy was bought on Steam (versus key-activated).
  • Received for free — true if the copy was a free key, useful for filtering out review-event noise.
  • Written during early access — true if the review was posted while the game was still in early access.
  • Created at and Updated at — original timestamp and last-edit timestamp in UTC.

Common workflows

  • Review-bomb forensics — group by day, plot volume, and overlay your patch and PR calendar. Coordinated spikes show up almost cartoonishly fast once the chart is in front of you.
  • Refund-window filter — filter to playtime_at_review_minutes > 120 so you're only looking at people who passed Steam's two-hour refund threshold and stuck around.
  • Localization triage — pivot by language and recommendation. If your global score is 82% positive but Simplified Chinese is sitting at 41%, that's a translation or culturalisation problem hiding in plain sight. Persona 5 Royal and Metaphor: ReFantazio both saw this kind of language-by-language gap on launch.
  • Patch impact — split the dataset into windows around a major update and compare percentage-positive plus average review length before and after. Length tends to spike when people are angry, which is its own signal.
  • Helpful-negative mining — sort by helpful_votes within the Not Recommended set. The top of that list is the criticism your community has effectively endorsed, and it's usually the most actionable feedback you'll read all quarter.
  • Free-key vs paid signal — split by received_for_free to see whether keys handed out at launch are pulling the headline number up or down.

Plan limits and API access

The Free tier returns up to 100 reviews per export — enough to confirm the format works for what you need. Personal scales to 5,000 results per export, Premium to 50,000, and Business to 250,000, which is enough to capture every review for the largest titles on Steam. If you'd rather pull reviews on a schedule or kick off an export from your own pipeline, the same job is available through the REST API and via webhooks. Full breakdown on the pricing page.

FAQ

  • Does this work for early-access games?
    Yes. Reviews written during early access are included and flagged in the written_during_early_access column so you can filter or weight them separately.
  • Can I export only English reviews?
    Yes. You can either restrict at export time or pull every language and filter on the language column in Excel — most analysts prefer the second option because it keeps the full dataset for later.
  • Are off-topic or moderated reviews included?
    The exporter pulls Steam's public review feed, which mirrors what's visible on the storefront. Reviews Steam has hidden as off-topic activity are excluded by Steam itself.
  • Why are there two playtime columns?
    Steam stores both the playtime at the moment the review was written and the player's current lifetime playtime. The first is the historical snapshot; the second tells you whether the reviewer kept playing.
  • Can I schedule a weekly export?
    Yes. Scheduled exports are available on Premium and Business — useful for tracking sentiment shifts week over week without re-running the job manually.
  • What if I have hundreds of titles to export?
    Use bulk mode: paste one Steam URL per line and the run returns one file per URL packaged in a single ZIP, so each title's data stays cleanly separated.