Product audit

Every check Product Pelican runs on your products, and how to fix each one.

The Products page is where most of the work happens. It reads your active products from Shopify, runs 14 checks against each one, and groups the failures into tabs you can work through.

📸 Screenshot needed — the Products audit page with the check tabs across the top, counts visible, and a product list below.

How the page loads

Large catalogues stream in. The page is usable as soon as the first batch of products arrives and continues filling in behind you, with a progress indicator showing how many of the total have loaded.

Results are cached between visits so the page paints instantly rather than re-reading your catalogue every time. The cache refreshes automatically after a full load; you can also force a fresh read from the page.

Only active products are audited by default. Draft and archived products are excluded. Widen the status filter on the page if you need them included — note that draft products are not loaded up front, so the page fetches them on demand.

Your plan caps how many products are audited. On a capped plan the catalogue is truncated and the page tells you so. See Plans & billing.

The checks

Each check has a severity: critical (red) for something actively broken, warning (amber) for something incomplete.

Categories

Check	Severity	Flagged when
No Category	Warning	The product has no Shopify category, or it is set to "Uncategorized"
Missing Category Attributes	Warning	The category is set but its attributes (Colour, Size, Material, …) are not filled in

Shopify's native product category drives a lot: it tells marketplaces, ad platforms and AI systems what kind of thing your product is, and it determines which attributes are available to fill.

Fixing it. Select products and generate. Product Pelican searches Shopify's product taxonomy for candidates and picks the best match, using the product title, type and tags. Category suggestions are constrained by product type — a knitwear product cannot end up filed under "Home & Garden" because of a fuzzy text match.

If you have Google Shopping fields switched on, accepting a category also sets the matching Google product category — Shopify's taxonomy already maps one to the other, so it is looked up rather than guessed, and the suggestion list shows it before you accept.

Once a category is set, Missing Category Attributes becomes fixable. The app reads the attribute list for that specific category and fills the choice-based ones from your product data.

Set categories before filling attributes. Attributes are defined per category, so there is nothing to fill until the category exists. This is also why an uncategorised product scores zero on two pillars, not one.

Titles

The Titles tab covers two checks with three presets:

• Empty — the title is blank

• Short — the title is under 20 characters

• All — every product, for rewriting titles that aren't technically failing

Titles are also flagged as vague when they match obvious placeholder patterns — Product 1, New product, Untitled, Test, Item 4, Sample, Draft.

Fixing it. Generate suggestions and review. A good product title names the product type and its key distinguishing attribute, and reads as something a person would search for.

Descriptions

Three presets:

• Empty — no description at all

• Short — under 100 characters of actual text (HTML tags don't count)

• All — every product

Descriptions are separately flagged as ambiguous when they lean on filler with no concrete subject — phrases like "the item", "this product", "click here", "learn more", "find out", "our product". These read as padding to a human and carry no information at all to a machine.

Fixing it. This is the most capable generator in the app. Alongside the description it can also:

• fill metafields you've configured, in the same pass

• read your product photos, if image analysis is on

The description and each metafield have their own accept checkbox, so you can take the metafields and reject the description, or vice versa. Each has a Show thinking link exposing the reasoning behind it.

📸 Screenshot needed — the Descriptions drilldown showing a generated description, generated metafield rows with individual checkboxes, and the Show thinking links.

SEO

One tab covering both the SEO title and the meta description, with two presets: Missing and All.

Shopify falls back to the product title and description when these are empty, so your pages still work — but nothing on the page has been written for a search result. The SEO generator fills both fields together.

Images and alt text

Check	Severity	Flagged when
No Images	Critical	The product has no images
Only 1 Image	Warning	Exactly one image
Missing Alt Text	Warning	One or more images have missing or poor-quality alt text

Alt text counts as poor, not just missing, when it is:

• under 20 characters

• a filename-style string with no spaces (blue_shirt_front_02)

• a generic word — image, photo, picture, img, product, item, untitled, or just a number

Fixing it. Alt text is generated per image. With image analysis on, the AI looks at the actual photo. With it off, it works from the product data and clues in the filename — it recognises conventions like front, back, detail, lifestyle, flatlay, model, hanger, swatch and describes accordingly.

The No Images and Only 1 Image checks can't be fixed by generation — they need real photography. They are surfaced so you know which products are affected.

Variants

Check	Severity	Flagged when
Missing SKUs	Warning	Any variant has no SKU
Missing Barcodes	Warning	Any variant has no barcode
Zero Price Variants	Critical	Any variant is priced at 0.00

These are reported, not generated — the app will not invent a SKU or guess a price. A zero-price variant in particular is usually a genuine mistake worth catching.

Product Pelican also notices cryptic option values — single- to three-character codes that aren't recognisable sizes or standard abbreviations. S, M, XL, 2XL, UK, EU and plain numbers are accepted; something like BLK or A1 on a Colour option is flagged, because a machine reading your catalogue cannot expand it.

Working through a tab

1. Filter the list. Presets narrow to a specific failure; you can also filter by status and tag.

2. Select products. Tick them individually, or select the whole filtered set.

3. Suggest Selected. The AI drafts a fix for every selected product.

4. Review. Read the suggestions, edit inline, untick anything you don't want.

5. Apply one product, or Accept Selected to write them all. Bulk writes ask for confirmation first.

Generation never writes to Shopify on its own. Suggestions live in the page until you accept them. Navigating away discards them.

Jumping straight to a product

The Enrich with Product Pelican action on the Shopify product page (under More actions) opens the audit page with that product already selected and scrolled into view, on the Descriptions tab.

If the product isn't in the currently loaded set — because it's a draft, or because your plan cap truncated the catalogue — the page widens the status filter automatically and tells you if it still can't find it.

Related pages

• GEO score — how these checks roll up into a score.

• AI prompts — control how each generator writes.

• Collection audit — the same idea, for collections.