Amazon URL and ASIN Import: What Graflio Reads From a Listing

Paste a ten-character ASIN or any Amazon product URL and Graflio reads the live listing before it draws anything: title, brand, bullets, the specification table, price, star rating, the reviews Amazon embeds on the detail page, up to twelve gallery photos, the colour variations and the category best-seller list around it. That evidence becomes a written brief, and the brief becomes the image set. This page is the exact inventory of what is read, what is not, and where the chain breaks.

What you can paste

Three input shapes work: a bare ASIN, a listing URL in any form Amazon serves it (/dp/ASIN, /gp/product/ASIN, the mobile /gp/aw/d/ASIN), or the tracking-laden link you get when you share a page off your phone. Referral parameters are stripped and the request rebuilt against the canonical /dp/ASIN address, because those long URLs are themselves a bot-check trigger.

The marketplace comes from the domain, and amazon.es, .de, .fr, .it, .co.uk and .com.mx are requested with a matching language header, so the listing arrives in its own language and the brief and headlines come back in that language rather than through a translation step. You can paste a list rather than one item; submissions are deduplicated on ASIN plus marketplace, and an account holds up to 200 products.

What is read from the listing

What is not read, and why

The full review corpus is not available to anyone. Amazon's dedicated /product-reviews pages sit behind a login wall, so the detail page's embedded selection is the ceiling: eight to ten reviews per listing, chosen by Amazon. A listing with 4,000 ratings does not hand the brief 4,000 reviews. It hands over the same handful a shopper sees, plus the same handful from each competitor, which is where the volume actually comes from.

Also not read: customer questions and answers, existing A+ content, the buy-box seller and fulfilment method, inventory, and any variation axis other than colour, so a size or pack-count family reads as one product. There is no advertising or search-term data in the pipeline, because none of it is on the public page.

The structural point is that this is a public-page read, not an account integration. Graflio never asks for Seller Central credentials and holds no SP-API connection. Nothing is written back to your listing and nothing needs your permission, which is also why the import works on a competitor's ASIN exactly as well as on your own.

The competitor pass

Once your listing is parsed, the deepest category from its Best Sellers Rank block is used to fetch that category's best-seller list. Page one carries thirty items, and up to ten of them, excluding your own ASIN, are then read as full listings in their own right: title, brand, price, rating, rating count, eight bullets, ten reviews and the main photo each, roughly a second apart. If your ASIN appears in the list, its rank is recorded.

Ten is a deliberate ceiling. Twenty competitors produced richer pain points but left the language model so little room that it regularly ran out of output tokens mid-brief. If one competitor page fails it is skipped; if the best-seller lookup fails entirely, research continues on your listing alone and tells you so.

From scraped data to a brief

Everything above goes into a single analysis pass. What comes back is structured, not prose: four to seven pain points, each tagged as coming from your reviews, a competitor's or both, with the evidence attached and the competitor named by brand; four to six USPs grounded in praise in your reviews, complaints in theirs, or a measurable spec, price or rating gap; one positioning sentence for the set; a visual style as hex colours plus a headline and body typeface, which a brand kit overrides; and six or seven image scripts, each with an exact headline, two to four callouts and a composition, varied across feature callouts, comparison, social proof, lifestyle, problem-to-solution and spec sheet.

One rule is enforced hard here: every statistic, percentage, review count, price and guarantee in the brief must appear verbatim in the scraped data. If the listing does not contain a number, the brief may not produce one. The whole pass costs 0.1 credit, so researching a catalogue costs less than rendering two images.

Where it breaks

A non-Amazon URL is rejected immediately. A dead ASIN returns Amazon's own 404, which almost always means the link belongs to a different marketplace than the one you pasted. Rate limiting is the common failure: Amazon throttles hard, the fetch retries four times across rotating browser fingerprints, and if it still fails you are told to wait rather than shown half a product. Finally a page can load but not parse, when Amazon serves a layout variant the parser does not recognise; whatever was extracted is kept and you finish the details by hand.

Two limits beyond that. Price and rating are a snapshot taken at read time and never re-sync, so a set built in March still quotes March's price. And a pre-launch ASIN with no reviews starves the most valuable half of the analysis: the competitor pass still yields category-level pain points, but the "your own customers said this" column is empty.

When to skip the import

Use it when the listing is live, has reviews, and sits in a category with a working best-seller list. Enter the product manually when you are launching something not yet on Amazon, or when it is a bundle whose argument is not the individual item's. Skip the research step altogether if you already know what your six images should say: it exists to decide the argument, not to confirm one you have made.

Common questions

Can I import a competitor's ASIN?

Yes. The import reads the public product page, so any live ASIN works whether or not you own it. Sellers commonly research the category leader first, to see which pain points surface. What you must not do is reuse a competitor's photographs: the import reads their gallery for analysis, and the images Graflio builds are composited from the product photo you supply.

Which Amazon marketplaces work?

Any amazon.* domain is accepted. The Spanish, German, French, Italian, UK and Mexican marketplaces additionally get a matching language header, so the listing is read in its own language and the brief comes back in it. Other marketplaces are read with an English header, which is fine for the data.

How many reviews does it actually read?

Eight to ten from your listing and up to ten from each of up to ten competitors, so around a hundred in a good run. Amazon's full review pages require a login, so nobody scraping public pages can promise more. A tool claiming to read thousands of reviews from a URL is describing something other than a public-page read.

Does it connect to Seller Central?

No. There is no account link, no API credential and no write path back to Amazon. Graflio reads public product pages and gives you files to download; you upload them to Seller Central yourself. That is a real limitation next to tools that publish directly, and it is also why nothing in your account is exposed.

What if my listing has no Best Sellers Rank?

The parser falls back to breadcrumb browse nodes, which resolve to a best-seller list for the broader category. If neither exists, the competitor pass is skipped and the brief is built from your listing alone: you still get USPs and image scripts, but the pain points are narrower because they can only come from your own reviews.