Walmart Scraper

Three surfaces on walmart.com, one actor: search returns products for a query or category URL, reviews returns a product's customer reviews, bestsellers returns a category's best-seller chart ranked by what Walmart sells.

No browser, no CAPTCHA solver.

What you get

Search mode

• Products from the search and category pages, about 60 per page, with rating and review count on nearly every row
• The barcode (UPC) on every detailed row. Walmart publishes it on the product page, and it is an exact join key against any other catalogue: match a Walmart listing to a Beslist, Discogs or Amazon record without fuzzy title matching
• Price, brand, seller, category path and description, all from each product's own page. Walmart serves none of them in search results, so they come with the detail stage
• Prices that belong to the right product. A Walmart product page carries dozens of price blocks: variants, bundles, "similar items" and sponsored rails. This actor anchors on the product's own id and refuses to reach beyond a bounded window, so a row either gets its own price or none at all
• was_price where there is a genuine markdown, filtered so it must be above the current price
• Sponsored placements flagged rather than mixed in silently
• price_source on every priced row, saying whether the number was found next to this product's id or elsewhere on its page

Reviews mode

• Reviews with the whole field set a reviews API is expected to carry: star rating, title, body, author, date, the Verified Purchase flag and both helpful counts, plus reviewer id, seller, fulfilment and photo count
• The rating histogram and Walmart's own totals on every run: how many ratings, how many written reviews, the 1-to-5 split and the average
• All six sorts the site offers, wired server-side and confirmed by the sort the page echoes back: relevance, most helpful, newest, oldest, highest rating, lowest rating
• max_reviews as a hard cap, 1 to 200. Walmart serves 10 per page
• A star filter, applied over the pages fetched, because Walmart filters stars in the browser only

Bestsellers mode

• The chart for a category, ranked by Walmart's own best-selling order: rank, item id, title, price, rating and review count, plus brand, seller, availability and badge state
• Prices already on the rows. Unlike search results, the best-seller listing carries a real price on every tile
• The chart title and the sort the page applied, both in the run summary, so you can see which chart you got
• Rank continuing across pages, 1 to 600 rows

What to know before you run it

Walmart answers HTTP 200 to everything, including to requests it refuses, where the body is a 15 KB "Robot or human?" page. Judging this site by status code alone says it is wide open; judging by size alone says it is closed. Both are wrong, which is why this Actor scores success on the product data itself.

The search payload has its prices removed. A successful request returns the full product list with names, brands, ratings and review counts, and price: 0 with an empty price block on every single product. This is Walmart's choice, not a parsing failure, and it is invisible if you count how often the word "price" appears in the JSON: the keys are all present, the values are not.

That is why prices are a separate, opt-in stage. It is one request per product instead of one per sixty, so a priced run is far slower and costs more. Making that a checkbox rather than doing it silently is the point.

"No reviews" and "we were not served the reviews page" are different answers. A product with no reviews returns a well-formed page whose review count is zero; a refused request returns the bot wall. This actor reports them separately: an empty answer is a successful run with zero rows and outcome: no_reviews (or empty_chart for a chart), while a refused or unrecognisable page is outcome: blocked with the code bot_wall or no_payload, and the run fails. An unknown category id is a third case again: Walmart answers 200 with a home page and no listing, which is reported as not_found. Nothing here returns an empty list and a success flag on a page it could not read.

Input

Field	Type	Required	Default	What it does
mode	select	no	search	search, reviews or bestsellers
query	string	search	laptop	What to search for
url	string	one of the two	–	A Walmart URL. Search or category in search mode, a product page in reviews mode, a category or chart in bestsellers mode. Takes precedence
product_id	string	reviews	–	Walmart item id or product URL. Wins over url
category	string	bestsellers	electronics	Department name, Walmart category id, or category URL
sort	select	no	relevancy	Reviews only: relevance, most helpful, newest, oldest, highest rating, lowest rating
rating_filter	integer	no	–	Reviews only: keep only 1-to-5 star reviews
max_reviews	integer	no	30	Reviews only: reviews to return, 1 to 200
fetch_prices	boolean	no	false	Search only: open each product's page for price, UPC, brand, seller, category and description
limit	integer	no	60	Search and bestsellers: rows to return, 1 to 600
retries	integer	no	3	Retry attempts per page

Output

Search mode, one row per product:

{
  "position": 2,
  "item_id": "18656507313",
  "url": "https://www.walmart.com/ip/HP-14-inch-HD-Windows-Laptop/18656507313",
  "title": "HP 14 inch HD Windows Laptop Intel Processor N150 4GB 128GB UFS Waterfall Blue",
  "rating": 4.1,
  "reviews_count": 805,
  "availability": "IN_STOCK",
  "sponsored": false,
  "price": 259.0,
  "was_price": 379.0,
  "price_text": "$259.00",
  "currency": "USD",
  "price_source": "product_page",
  "query": "laptop"
}

Reviews mode, one row per review:

{
  "position": 2,
  "review_id": "432855292",
  "rating": 5.0,
  "title": "Worth the purchase!",
  "text": "This computer came packed with everything you'll need. It has a camera with the lense blocker already installed. I also have a 1 year Microsoft office subscription, that I formally had to buy separately when purchasing a computer. Modern programs, windows and apps. It also has lots of storage/GBs. Finally, my Favorite thing is that it's Pink of course!",
  "author": "Keyerra",
  "author_id": "2cb099fc76d45d4c25fbb56041abe37cd19cadb2e571d9cce05a73fa7c6b6632bff27bf9e6de41a1131d47cf01a241ae",
  "date": "7/17/2026",
  "verified_purchase": true,
  "helpful_positive": 0,
  "helpful_negative": 0,
  "seller_name": "Walmart.com",
  "fulfilled_by": "Walmart",
  "item_id": "18611919209",
  "status": "APPROVED",
  "product_id": "18656507313",
  "sort": "relevancy"
}

Bestsellers mode, one row per chart entry:

{
  "rank": 1,
  "item_id": "17771711203",
  "url": "https://www.walmart.com/ip/Open-Box-Restored-Apple-iPhone-14-Pro-Fully-Unlocked-128-GB-Space-Black/17771711203?conditionGroupCode=3&classType=REGULAR",
  "title": "Open Box Apple iPhone 14 Pro 128GB Space Black LTE Cellular 3L242LL/A",
  "brand": null,
  "price": 362.0,
  "price_text": "$362.00",
  "currency": "USD",
  "price_source": "listing",
  "rating": 1.0,
  "reviews_count": 1,
  "seller": "Walmart",
  "availability": "IN_STOCK",
  "best_seller_badge": null
}

On provenance: the review and chart rows above are real captures, not illustrations. The review is row 2 of tests/fixtures/reviews_hp_laptop.json (2026-09-24) and the chart entry is rank 1 of tests/fixtures/bestsellers_electronics_page1.json (2026-09-25, https://www.walmart.com/browse/electronics/3944?sort=best_seller), so every figure in them can be re-read from the repo. The search row is the field shape only; no search capture is committed. Note the nulls in the chart row: Walmart left brand and best_seller_badge empty on that tile, and the parser keeps them empty rather than guessing.

The run summary carries the outcome on every run: ok, no_reviews, no_matching_reviews, empty_chart, not_found or blocked, plus the counts behind it.

Use cases

Catalogue and assortment work, cheaply. Leave fetch_prices off and a run is one request per sixty products. You get the full product list with ratings and review counts, which is enough to see what Walmart stocks in a category and which items have traction.

Price monitoring on a shortlist. Turn fetch_prices on for a narrow query rather than a broad one. Fifty tracked products is fifty requests, which is a sensible daily job; five thousand is not.

Finding what people actually buy. reviews_count beside rating separates a 4,8 with eleven reviews from a 4,1 with eight hundred. The second is the one that sells.

Marketplace seller research. seller shows where a listing is Walmart's own and where it is a third party, which is the same question the Amazon buy box answers.

Watching promotions. was_price is only set when there is a real markdown, so a scheduled priced run on a small set is a discount alert.

What buyers actually say. Reviews mode gives you the review text and the Verified Purchase flag, so you can tell a real complaint from an unverified one. helpful_positive sorts the ones other shoppers found useful.

Following sentiment over time. sort=submission-desc with a scheduled run is a drip of what people said this week about a product, at ten reviews per request.

Assortment against the market. Bestsellers mode is the ranked answer to "what is selling in this category", with price and review count beside each rank, which is the comparison a buyer makes.

How it compares

	this actor
Per 1.000 products	$1,00
Per 1.000 resolved prices	$3,00, and only for prices actually found
Needs a browser	no
Prices anchored to the right product	yes
Sponsored placements flagged	yes

Most published Walmart actors reach the site through a browser or a paid unblocking service, because the plain request returns a bot wall. This one uses a crawler identity that Walmart still serves, which is why it is fast and cheap, and which is also its main risk: single-identity routes close. If it does, the run reports bot_wall and explains it rather than returning nothing.

Pricing

Three events. run_start costs $0,0010 per run. product costs $0,0010 per row written, which is $1,00 per 1.000, and that is the unit in all three modes: a review row and a chart row bill the same as a product row. product_detail costs $0,0030 and is charged only when a product page was actually read in search mode, never for a lookup that came back empty. A run that returns nothing is billed run_start alone.

Limits and gotchas

• No prices without fetch_prices. On search rows only. This is Walmart's doing. The run summary says so on every unpriced run rather than leaving an empty column unexplained.
• The price comes from the product's own object, not from nearby text. An earlier build searched forward from the product id for the closest price block and kept finding the "similar items" rail: it returned $949,99, then $539 on the next run, for a Chromebook that was neither. Reading priceInfo off the object already matched on usItemId cannot pick up a neighbour's price. Prices are now stable run to run.
• Priced search runs are slow. One request per product, deliberately unparallelised. Sixty products takes minutes; six hundred takes a long time. Set the actor timeout accordingly.
• brand is sparse because Walmart often omits it from the search payload.
• Sponsored rows are included, flagged with sponsored, because filtering them out silently would misrepresent what the page shows.
• The star filter is not server-side. Walmart's rating histogram is a browser control and every server-side star parameter we could find is ignored. rating_filter is therefore applied to the pages fetched, which is enough for most products and can stop early on a long-tailed one; ratingFilterMode and reviewsScanned in the summary say how far it got.
• Best-seller ranks are only ranks when the page says so. The chart is Walmart's sort=best_seller order. Some Walmart URLs, notably the /cp/best-sellers/ landings, ignore that sort and serve a merchandising shelf instead; the summary reports sortEcho, sortApplied and chartNote so a shelf is never presented as a sales ranking.
• US only.

FAQ

Why are all my prices and barcodes empty? Because fetch_prices is off. Walmart's search payload has no price, no UPC, usually no brand and no category path. All of it lives on the product page, so it costs one request per product.

Can I get the barcode? Yes, upc, with the detail stage on. It came back on every row in testing.

Why did some products get no price even with fetch_prices on? Because no price block was found close enough to that product's own id on its page. The actor reports that as absent rather than substituting a nearby number that may belong to a different product.

Does it need a proxy or a browser? A residential US exit, yes. A browser, no.

How many products can I get? Up to 600. Without prices that is ten requests; with prices it is 610.

My reviews run says "no reviews". Is that real? Read outcome in the summary. no_reviews means Walmart served the reviews page and it reports zero; blocked means we were given the bot wall or a page without the reviews payload, which is a failed run, not an empty one. A product can also have ratings and no written reviews, and the message says that too.

Why is best_seller_badge sometimes missing rather than false? Because Walmart did not say. A tile with no badge block is unknown, so the field stays empty instead of claiming the product is not a best seller.

How do I know I got the real best-seller ranking? The summary's sortEcho is what the page applied. sortApplied: true means the order is Walmart's best-selling order. If it is false, chartNote says what the order is instead.

Related Actors

• Newegg Scraper — US retail with real discounts on the listing page
• Amazon Buy Box Monitor — who currently wins an Amazon listing
• Target Scraper — the other big US general retailer