Quick answer
How do AI WooCommerce upsells actually work?
Open an add-on group scoped to a single product and click Generate suggestions. The product's name, price and category are sent to the AI service, which proposes add-on ideas with a suggested price and a rough likelihood. Nothing is added to your store until you click Add to group. Treat it as a first draft that gets you past a blank screen, not as pricing advice.
The hardest part of selling add-ons is not building the fields. It’s sitting in front of a product you’ve sold for three years and having no idea what else a customer might want with it. AI WooCommerce upsells are useful for exactly that moment, and much less useful for everything people hope they’ll do.
This is what the feature does, what it sends, what it gives back, and where your judgement still has to do the work.
Where the panel appears
Open Smart Add-Ons → Add-On Groups and edit a group. The AI assist panel appears when the group is scoped to a specific product and that product is chosen. It’s deliberately not shown on global or category groups: the suggestions are generated from one product’s details, so there’d be nothing meaningful to work from.
If you don’t see it, check the group’s Scope is Specific product and a product is selected. AI suggestions are also a Pro-plan feature, so on a lower plan the panel explains that instead of offering the button.
What gets sent, and what doesn’t
Worth being precise about, because “AI plugin” makes people reasonably nervous about their data. Four things about the product are sent: its name, its price, its first category, and your site’s language so the suggestions come back in it.
Your orders, your customers and your sales history are not part of the request. Which is also why the suggestions are generic rather than tailored — the model is reasoning from “a 27-dollar T-shirt in Apparel”, not from what your buyers have actually purchased.
What comes back
Each suggestion arrives as a label, a short description of the idea, a suggested price, and a rough indication of how likely it is to convert. Each has its own Add to group button, and nothing reaches your store until you press one.
That one-click gap between suggestion and field is the whole design. The panel says “review before adding” because the model cannot know your margins, your suppliers, your lead times, or the thing you tried last year that generated forty support emails.
Using it well
The teams who get value out of AI WooCommerce upsells tend to use them in a particular way:
- Generate before you think, not after. The value is in breaking a blank page. Once you have your own strong idea, the list adds little.
- Keep one, discard the rest. A group with eight AI-suggested options is worse than one with two you believe in. Shoppers abandon long forms.
- Rewrite every label in your own voice. Suggested wording is generic by construction; your product page isn’t.
- Set your own price from your costs, then check it against the pricing guide for which price type fits.
- Measure it. A suggestion is a hypothesis. The analytics dashboard tells you whether shoppers actually picked it, and A/B testing tells you whether it beat what you had.
That last step is the one that turns AI WooCommerce upsells from a novelty into something useful. Generating ideas is cheap; knowing which of them earned money is the part that compounds.
What it won’t do
- It won’t add anything on its own. No background job creates fields. Every option on your store was put there by a person.
- It won’t learn from your store. Suggestions don’t improve as you sell, because sales data isn’t part of the request.
- It won’t check stock or feasibility. If it proposes gift wrapping and you don’t stock gift wrap, that’s yours to notice.
- It won’t write your conditional logic. Field relationships are still something you design.
Common questions
Does it need my own OpenAI key?
No. The request goes through the Dhyan service, authenticated with your licence, so there’s no separate AI account, key or bill to manage.
Will the same product always give the same suggestions?
Not necessarily — generated output varies. If the first set doesn’t fit, generating again is reasonable. If the third set still doesn’t fit, the product’s name and category probably aren’t describing it well enough, which is worth fixing for SEO reasons anyway.
Are AI WooCommerce upsells better than picking options myself?
For a product you know well, no. For the two hundredth product in a catalogue you inherited, it’s considerably faster than staring at it. Use it where you’re short of ideas, not where you’re short of confidence.
If you’re weighing up the plan, the Pro plan guide covers what else sits alongside it, and the upselling guide covers making the suggestions land without annoying anybody. For the wider picture on generated content and search, Google’s helpful content guidance (opens in new tab) is worth reading before publishing anything a model wrote.
Comments