> For the complete documentation index, see [llms.txt](https://native-app-co.gitbook.io/product-pelican/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://native-app-co.gitbook.io/product-pelican/changelog/readme.md).

# Changelog

New updates and improvements

{% updates format="full" %}
{% update date="2026-10-04" tags="feature" %}

## Jobs: Pelican finds the gaps and fixes them for you

Until now you opened the audit, picked a check and worked through it. **Jobs** turns that around: Product Pelican keeps checking your catalogue in the background, prepares each fix with your own AI prompts, and has it waiting for you.

Two jobs to start with:

* **Image alt text** — every image with no alt text gets one written for it.
* **Product categories** — every product with no category, or sitting in "Uncategorized", gets a category suggestion.

### You decide how much to hand over

Each job runs in one of two modes:

* **Ask me first** — fixes queue up on the **Jobs** page, and on your dashboard, for you to approve, edit or reject. Approve them one at a time or in bulk.
* **Automatic** — Pelican applies them on its own. Alt text is applied as it is written; categories are applied only when Pelican is highly confident, and anything less certain still waits for you.

Once you've approved a run of fixes with almost no changes, Pelican offers to take that job over automatically.

### Safe by design

* **Jobs only fill blanks.** They never rewrite alt text or a category you already have.
* **Everything can be undone.** The **Activity** tab keeps the last 14 days, with undo for a single fix or a whole day.
* **Rejected means rejected.** A fix you turn down is not suggested again.

| Plan                      | Fixes prepared per job, per day |
| ------------------------- | ------------------------------- |
| Standard                  | 100                             |
| Pro (and the 7-day trial) | 300                             |
| Max                       | 1,000                           |

### This is just the beginning

Alt text and categories are the first two jobs, not the last. We will be adding more automated tasks, and we can also build custom tasks for your store on request — tell us what you'd like Pelican to take off your plate.

{% hint style="info" %}
**Jobs is rolling out store by store.** If you don't see **Jobs** in the menu yet and want it now, contact support and we'll switch it on.
{% endhint %}
{% endupdate %}

{% update date="2026-10-04" tags="improvement" %}

## Image analysis: watch your product photos become product content

{% embed url="<https://youtu.be/Xr1bcljr3k8>" %}

A new walkthrough of image analysis, start to finish: Product Pelican looks at your product photos and writes from what is actually in the shot, not just from the title and tags.

One switch, and your photos feed all of this:

* **Product descriptions** — colour, material, cut, pattern, styling details and what is visible in the photo.
* **Your own metafields** — every metafield you've set up to generate alongside the description is filled from the same photos, so attributes like colour, material or finish stop being guesses.
* **Google Shopping fields** — condition, gender, age group, custom product and custom labels 0–4, ready for your Google feed.
* **Image alt text** — written from the image itself rather than from its filename.

Check up to **three photos per product** when a second or third angle shows the back, the fabric or the scale.

Turn it on at **AI Prompts → Description → Image analysis**. It is off until you do, and each analysed photo counts once towards your plan's monthly allowance.

{% content-ref url="/spaces/gsM5MoxEK42f38xn0qO8/pages/yjWKoB72ixJVMpHGb950" %}
[Image analysis](/product-pelican/ai-content/image-recognition.md)
{% endcontent-ref %}

[Image analysis](/product-pelican/ai-content/image-recognition.md)

## Write an FAQ once. Publish it across the whole product family.

{% embed url="<https://youtu.be/-sSn3Y2_V08>" %}

A jacket in six colours is six products — and until today, six rounds of adding the same FAQs. Now it is one.

Add an FAQ to a product and Product Pelican puts it on every related product in the same click. Ten colourways, a matching set, a full size run: answer the question once and the whole family is covered, structured data and all.

### How it works

1. On **AI prompts → FAQ Generation**, choose a **Product grouping** — any product metafield that links to other products. If you use Product stitching, the links are already there.
2. Open a product on the **Product FAQ** page. Its related products appear as a row of thumbnails under **Also apply to related products:**, all selected.
3. Add FAQs the way you always have — type one, accept the AI's suggestions, or connect one from your library. Each lands on every selected product at once.

Don't want a product included? Click its thumbnail and it sits that round out.

### Built to be safe on a live catalogue

* **One FAQ, not ten copies.** Every product in the group points at the same FAQ. Correct a returns policy once and it is corrected everywhere.
* **Nothing is overwritten.** Each product keeps the FAQs it already has, in the order you set. New ones are added to the end, and never twice.
* **You stay in control.** Deselected products are never touched, and if a product can't be updated the page tells you exactly which one and why.

Pair it with **Suggest FAQs** and one click of **Accept all** fills an entire product family with AI-written, structured-data-ready answers.

Product grouping is off until you choose a metafield, and applies to product FAQs.

{% content-ref url="/spaces/gsM5MoxEK42f38xn0qO8/pages/fajrrbKtUy9pMXzLpSIj" %}
[Product FAQ](/product-pelican/faq-generation/product-faq.md)
{% endcontent-ref %}
{% endupdate %}

{% update date="2026-09-26" tags="feature,improvement" %}

## 7-day Pro trial replaces the Free plan

Every new install now starts on a **7-day trial of the Pro plan** — everything Pro includes, on your full catalogue, with nothing charged and no card to enter. The Dashboard and the Plans page show how many days are left, and the Pro card says **Activate Pro** whenever you're ready to carry on without interruption.

When the trial ends and you haven't picked a plan yet, the app asks you to choose one before you continue. You are never charged automatically.

**If you were on the Free plan:** the Free plan is gone. Shops installed before today are treated as having already used their trial, so you'll see the choose-a-plan prompt on your next visit. Nothing you generated is lost — descriptions, FAQs, categories and metafields all stay on your products in Shopify.
{% endupdate %}

{% update date="2026-09-09" tags="feature" %}

## Pelican Brain

Product Pelican now researches your store the moment it is installed. A minute later, **Pelican Brain** — new in the app navigation, with a banner on the Dashboard — shows you what it found:

* **What we found** — your catalogue size, typical price range, main categories and the brands you carry.
* **Your key products** — your best sellers, weakest first, each with its GEO score and the data gaps holding it back. **Fix** takes you straight to the right audit check.
* **AI Visibility Check** — whether AI assistants recognise your brand at all, how one would describe your store from your public pages, and what it cannot answer (shipping, returns, notable products). Anything not findable on your site is marked as such, never guessed.
* **Suggested prompts** — the product description, collection description and FAQ prompts, each pre-written in your brand's voice on top of the standard prompt. Edit, tick and **Apply selected**; they show up on AI Prompts as *Customized*.
* **Likely FAQ topics** — the questions shoppers of a store like yours tend to ask.

{% hint style="info" %}
**Pelican Brain never changes your store on its own.** It only reads your catalogue and public pages. Your AI prompts change only when you apply a suggestion. Re-run the research any time from the top of the page.
{% endhint %}
{% endupdate %}

{% update date="2026-08-24" tags="fix" %}

## Descriptions no longer cut off

Fixed a bug where descriptions generated with product images enabled could come back empty or stop mid-sentence: the AI's reasoning was using up the space reserved for the answer. The AI now has enough room to think *and* write the full description, and a cut-off answer is retried automatically rather than shown as a suggestion.

Also fixed: choosing **4 or 5** images per product silently disabled image analysis (the image model only accepts three) while still counting the images against your monthly allowance. The setting now goes up to **3**; shops that had 4 or 5 selected are treated as 3.
{% endupdate %}

{% update date="2026-08-18" tags="fix" %}

## More reliable description generation

Fixed an issue where generating a description with product images enabled could occasionally end with a "No suggestion returned" error even though the AI had done the work. Generation now automatically retries, so descriptions come back reliably on the first click.
{% endupdate %}

{% update date="2026-08-18" tags="feature" %}

## Google Shopping fields

{% embed url="<https://youtu.be/_e72l3qzHog>" %}

{% content-ref url="/spaces/gsM5MoxEK42f38xn0qO8/pages/meCFzrqHQjKk4ELidGZL" %}
[Google Shopping fields](/product-pelican/products-and-collections/google-merchant-center.md)
{% endcontent-ref %}

Product Pelican can now fill the **Google Shopping field set** on your products — the data your Google feed is built from. Category, condition, gender, age group, custom product and five custom labels, written where every feed tool already looks for them.

Find it on **AI Prompts → Category** (for the Google product category) and **AI Prompts → Metafields → Google Merchant Center** (for everything else).

### The Google product category is looked up, not guessed

Shopify's product taxonomy already carries a matching Google category for every category it knows. So Product Pelican doesn't ask the AI twice — it sets the Shopify category, then reads the Google one straight off the mapping.

Switch it on and accepting a category suggestion sets both. The suggestion list shows the Google category underneath each candidate, so you see what will be written before you accept it. Products that already have a category pick theirs up in the **Descriptions** check, so there is no need to re-categorise a catalogue that is already done.

### The rest fills alongside your descriptions

| Field            | What goes in it                                                       |
| ---------------- | --------------------------------------------------------------------- |
| Condition        | new · refurbished · used                                              |
| Gender           | male · female · unisex                                                |
| Age group        | newborn · infant · toddler · kids · adult                             |
| Custom product   | true for handmade, made-to-order or own-brand items without a barcode |
| Custom label 0–4 | Whatever you segment your campaigns by                                |

Each one gets its own row and its own accept checkbox in the results, and the fields with a fixed vocabulary are shown as a dropdown — the AI cannot invent a value Google would reject.

{% hint style="info" %}
**The fields appear in your Shopify admin too.** The first time a value is written, Product Pelican creates the field's definition in your store, so you can see and edit it on the product page. If you already have one, yours is left untouched.
{% endhint %}

{% hint style="warning" %}
**Nothing is written until you turn it on.** Every Google field is off by default. Gender and age group in particular are meant for apparel — leave them off for homeware, tools or food.
{% endhint %}
{% endupdate %}

{% update date="2026-08-11" tags="feature" %}

## Image analysis

Product Pelican can now **look at your product photos**. Turn image analysis on and your photos are shown to the AI alongside the product data, so what it writes is based on the product as it actually appears — not just its title, type and tags.

Available to all stores, on every plan. Find it on **AI Prompts → Description**, under **Image analysis**.

{% embed url="<https://youtu.be/BqQOzYAfd78>" %}

### [Image analysis](/product-pelican/ai-content/image-recognition.md)

### What it improves

Once it is on, photos feed three things:

* **Product descriptions** — colour, material, cut, pattern, styling details and what is actually in the shot.
* **Generated metafields** — the fields generated alongside the description are grounded in the same photos, so attributes like colour or material stop being guesses.
* **Image alt text** — alt text is written from the image being described rather than from filename clues, which is what makes it genuinely useful for accessibility and SEO.

### Choose how many photos to check

Set **Images checked per product** to anywhere from **1 to 5**. More photos give the AI a fuller view — useful for products where a second or third shot shows the back, the fabric, or the scale — at the cost of more of your monthly allowance.

Alt text always uses the single image it is writing for.

### Your monthly allowance

Each analysed photo counts once against a monthly limit that depends on your plan.

| Plan     | AI image analyses / month |
| -------- | ------------------------- |
| Free     | 10                        |
| Standard | 1,000                     |
| Pro      | 3,000                     |
| Max      | 6,000                     |

The allowance resets on the 1st of every month.

{% hint style="info" %}
**Running out never blocks you.** If you use your allowance up, generation carries on exactly as before — from the product data, without the photos — and the description panel tells you the images were skipped. Nothing fails and nothing is lost.
{% endhint %}

{% hint style="warning" %}
**Image analysis is off until you turn it on.** No photo is analysed, and nothing counts against your allowance, until you tick **Use product images in AI generation** yourself.
{% endhint %}

<details>

<summary>Also in this release</summary>

* **See the AI's reasoning.** A **Show thinking** link now sits next to the generated description and next to each generated metafield, revealing the reasoning behind the suggestion — including on generations that fail, so you can see what went wrong.
* **Accept the description and the metafields separately.** The description has its own checkbox, so you can apply only the metafields you want without touching the description.
* **Missing metafields are called out.** If a configured field comes back empty, it is listed rather than silently dropped.

</details>
{% endupdate %}
{% endupdates %}