Can Claude upload Shopify products automatically from supplier PDF catalogue?

Technical Q&A
, ,
1

Hey guys, I’m wanting to know if Claude and upload products to my store? I have multiple suppliers who provide their product catalogues, specifications and wholesale price lists as PDF files. Some of these documents contain hundreds of products, so manually creating or updating each Shopify product one by one is extremely time-consuming.

I’m trying to understand whether Claude can help automate this process.

Ideally, I would like to upload the supplier’s PDF catalogue and price list to Claude and have it:

  • Extract the product names, descriptions, SKUs, specifications, variants and pricing.
  • Match the catalogue information with the correct prices.
  • Format the information correctly for Shopify.
  • Either create the products directly in Shopify or generate a Shopify-compatible CSV that I can review and import.
  • Update existing products without creating duplicates.

I’m also unsure how product images would work. Most supplier catalogues contain at least one image for each product.

Can Claude:

  • Extract the product images directly from the PDF?
  • Download higher-quality images from the supplier’s website?
  • Match each image to the correct product and variant?
2

Hi @alxure

Yes, this is definitely possible.

Claude can help extract product information from your supplier PDFs, including product names, descriptions, SKUs, specifications, variants, and pricing. It can also match data from separate catalogues and price lists, then format everything into a Shopify-compatible CSV for review and import.

Regarding images:

If the images are embedded in the PDF, they can be extracted and matched to the correct products.
If the supplier has higher-quality images available on their website, we can also retrieve those and associate them with the correct products and variants.

The exact approach depends on the quality and structure of your supplier files, but this type of workflow can significantly reduce the manual effort involved in managing large product catalogues.

Thanks.

3

Yes it can. However, you’re going to spend a fair amount of time and money getting a full product catalog in PDF form into viable Shopify products, let alone be able to accurately match those products up with a price list in excel/csv format. Not against this at all (FyreTrail does it seamlessly with Purchase Orders), but being the developer of FyreTrail I can tell you there are quite a few gotchas to watch out for and PDF size is most certainly one of them. You might be working at this for a while, but you can do it if you have the time to crank it out.

4

Yes, Claude can handle most of the heavy lifting here with Claude Cowork.

Cowork can read and write files in a folder you grant it access to, run multi-step tasks autonomously, and pair with the Claude in Chrome extension to fetch things from the web.

What you need to prepare:

  • A paid Claude subscription (Pro, Max, Team, or Enterprise)
  • Claude Desktop app installed (Windows or macOS)
  • Claude in Chrome extension installed

How to start:
Open Claude Desktop → switch to Cowork mode → point it at the folder with your supplier PDFs → paste your question there + ask it what it can and can’t do for your specific setup, and it will walk you through the rest.

image
image883×927 116 KB

Best,

Ellie from BOGOS: Free Gift Bundle Upsell team

5

Hi thanks for your reply. yeah the photos are on the supplier website

6

You can use the images from the supplier’s website by matching each product using its SKU or product code. Once a product is identified, the system can fetch the corresponding image URL from the supplier’s website and include it in the Shopify import, or upload it directly if we’re using the Shopify API.

So instead of using the lower-quality images from the PDF, your Shopify store will use the higher-quality images available on the supplier’s website.

7

Hey @alxure

hope you’re doing well!

Claude can definitely help extract product data from PDFs and generate a Shopify-compatible CSV for import. For direct product creation, updates and image handling, you’ll likely need to combine Claude with Shopify APIs or an automation tool to streamline the workflow.

8

is this something you can assit in? im unsure how to do it with the image handling.

9

Hey, I’ve just given it a go. However, a lot of the products that we uploaded were not the greatest quality-wise, and also in terms of sizing, they were quite small.

10

Yes, we can use the images directly from the supplier’s website, provided they are available in good quality. We can then update the image URLs in the CSV file and import them into your website, ensuring that the correct images are assigned to their respective products.

11

A few people already covered the “yes, Claude can generate the CSV” part, so I’ll focus on the two things that tend to bite when you import a few hundred AI-generated rows at once, plus the image sizing you mentioned.

On the images being small/low quality: that’s usually not a CSV problem, it’s the source. If the PDF-embedded images are low-res, matching each product to its higher-res image URL on the supplier’s site (by SKU or product code) and putting that URL in the image column is the fix. So the CSV just needs the right public https:// image URL, not the PDF image itself.

On the CSV itself, before you push all several hundred rows, I’d sanity-check the structure on one product group first, because AI-generated CSVs from PDFs tend to break in predictable ways:

  1. Group rows by handle and confirm the first row for each handle is the product row (title + product-level fields). Extra rows with the same handle should be variant/image rows, not a second product-start. AI extraction from a PDF often splits one product across rows inconsistently.

  2. Keep option columns consistent per handle. If a product has variants, every row for that handle needs the same Option1 name (e.g. “Size”, not “Size” on one row and blank on the next), and Option1/Option2 values must actually identify a unique variant. Duplicate option combinations under one handle can create or delete the wrong variants.

  3. Check every image value is a full public https:// URL. A bare filename, a PDF-relative path, or a login-protected supplier URL won’t import. Paste one URL into a private browser window; if the raw image doesn’t load without signing in, Shopify can’t fetch it either.

  4. Watch for rows that have image alt/position but an empty image URL, and for a stray column count mismatch — PDF-to-CSV steps love to shift a column and misalign the rest of the file.

  5. For “update existing products without creating duplicates”: match on the exact existing handle and use “Overwrite products with matching handles”. If the AI invented new handles, you’ll get duplicates instead of updates. And remember included blank columns can overwrite existing data, so only keep the columns you actually intend to change.

Import one handle first, confirm the variant count and images look right in admin, then run the full file. Catching a structural issue on 1 product is a lot cheaper than finding it after several hundred.

12

I did this by hand for a few suppliers before I got sick of it and built a tool, so a couple of things that bit me.

The bit that got me wasn’t the CSV format, it was the join. You’ve got catalogues and price lists as separate PDFs, so somewhere you’re matching a SKU in one doc to a price in another. When that goes wrong it doesn’t error, you just end up with the right product carrying the wrong price sitting live on your store. I’d spot check 20 random SKUs against the original price list before anything goes live, and pay extra attention to anything with tiered or wholesale pricing where there’s more than one number per row.

Other thing is layout. Supplier PDFs are built for a human to read, so you get multi column pages, price tables that break across a page boundary, and variant matrices with sizes running across the top and SKUs down the side. Extraction usually looks perfect on page 1 and quietly falls apart around page 40. Check the back of the file, not the front.

They also change the layout between versions, which is great fun when you’ve got multiple suppliers sending updates.

On the small images, agreed with what’s above, pull them from the supplier site by SKU rather than out of the PDF. Embedded catalogue images are usually downsampled to whatever the print spec needed.

Full disclosure, I built skuparse.com which does the supplier PDF and spreadsheet to Shopify CSV part, so I’m not neutral. Either way the spot check advice stands, that’s the one that saves you.

13

Yeah, Claude can help with this, but manage expectations a bit.

What it’s good at: reading your supplier PDFs and pulling out product names, SKUs, specs, and prices into a clean table, then formatting that into a Shopify import CSV (handle, Title, Variant SKU, Variant Price, etc). For a first pass on hundreds of SKUs, that alone saves you days.

What it’s not good at out of the box: it doesn’t have a live connection to your store, so it can’t push products in directly or check for duplicates against what’s already live. You’d need to export your current product CSV and feed it back in so it can match on SKU and avoid creating dupes.

Images are the weak point. It can pull the images embedded in the PDF. Still, catalogue images are often low-res (fine for print, rough for zoom-on-product-page). If the supplier has a website with the same products, better to source images from there instead.

Practical workflow: run one catalogue first, review the CSV output closely (especially price and variant mapping) before a real import, then scale up once you trust the accuracy. Don’t batch-import hundreds of rows unchecked on the first go. A mismatched price or wrong SKU-image pairing is an easy miss at that volume.

14

Treat this as a data-join and review workflow, not one long Claude prompt.

Keep a staging table with the source PDF, page, supplier SKU, supplier product URL, option values, price, image URL, and an extraction-confidence field. Normalize supplier SKUs, but do not use fuzzy title matching to decide a price or image; quarantine duplicate, blank, or conflicting SKUs for review.

Export the current Shopify catalog and build an exact Variant SKU map. Split the proposed output into:

  • new products;
  • exact-SKU updates; and
  • unresolved rows.

Use Shopify handles to group CSV rows, not as proof that a supplier item matches an existing variant.

For images, first confirm the supplier permits reuse. Then take the largest official product-image URL, record its pixel dimensions, and reject thumbnails or URLs that do not return an image content type. This avoids treating the small PDF preview as the source asset.

Pilot 10–20 products first. Check variant counts, option ordering, price joins, SKU/image pairs, and duplicate SKUs before scaling. After the import, re-export the pilot and compare it with the approved staging rows; that re-export is the evidence that Shopify stored what you intended.

One question changes the architecture: is this feeding one Shopify store or several, and can the same supplier SKU have different prices, titles, or images per store? If several stores intentionally differ, keep a canonical supplier layer plus explicit per-store overrides instead of asking Claude to regenerate each store independently.

15

Disclosure: I build a tool in this space, so weigh my answer accordingly.
Agreeing with the staging-table and pilot advice above — I’d add the layer that
actually decides whether the extraction is usable, which is the page structure,
not the model.
Four things break supplier catalogues, and each has a cheap rule:

  1. Anchor a record on the product CODE, never on a text line. One code = one row.
    A line with no code is either a wrapped continuation of the row above, or the
    description of the row below — decide by the vertical gap: tight gap = it
    belongs above, loose gap = it belongs below.
  2. Merged cells. One description is often printed once for a whole family and
    shared by 5 variant rows. It must be filled DOWN until the next family starts,
    otherwise four products lose their name. The family ends where the next family
    begins — not where the line ends.
  3. Page breaks. A product’s name sits on page 3 and its variants continue on
    page 4. Carry the last section/title across the page boundary or those rows
    arrive nameless.
  4. Prices and warranty: pull them by content type (currency symbol, “1 YEAR”),
    not by fixed column position — the columns drift between pages.
    The cheapest check that catches a bad read before import: count the product codes
    on the page with a regex, and compare with the number of rows you produced. If
    they don’t match, the read is wrong — no review needed to know that.
    On images: the pictures embedded in a PDF are usually the low-res print preview.
    Rendering the page at 300dpi and cropping the product’s region often beats the
    embedded file, and it needs no supplier permission question since it’s the same
    document you were given.
16

Using AI LLMs like Claude or ChatGPT to extract product tables from supplier PDFs is a great first step, but you’ll hit a hard technical wall when transferring those extracted tables to Shopify via CSV:

  1. PDF Image Extraction Gap: LLMs parse text tables well, but they cannot extract and host the embedded PDF image binaries. You end up with SKUs and prices, but blank image URLs.

  2. WebP and Image Parameter Traps: If the supplier’s PDF references external image CDN links (often raw WebP files from Asian manufacturers), Shopify’s CSV importer will drop them. Shopify strictly expects static JPG/PNG endpoints.

  3. Multi-Row Variant Mapping: A single PDF product with 4 colors × 3 sizes needs to be expanded into 12 CSV rows, with the color image URL explicitly duplicated across all size rows sharing that color.

If you’re sourcing from supplier catalogs/files and want to automate the end-to-end pipeline (data extraction + WebP-to-JPG local conversion + Matrixify-compliant CSV ZIP creation), consider using a local Chrome tool like EasyCatch. It runs 100% locally inside your browser sandbox, handling variant image mapping and format conversion without sending your supplier data to third-party servers.

17

One important distinction is whether the supplier PDFs are only a source for creating Shopify product data, or whether the original manuals, datasheets, certificates, or safety documents also need to remain available to customers after the products are imported.

The first workflow is a catalogue extraction and product-import problem. The second is an ongoing document-management problem, because those files may need product assignments, version updates, stable customer-facing links, and checks for broken source files.

I’d recommend treating those as two separate stages rather than trying to solve everything in one PDF-to-product import process.