If you've priced AI search for a WooCommerce store, you've hit the same two walls everyone does. Half the vendors don't publish a price at all. The other half say "free" — and they mean the plugin is free, not the AI behind it.
This is the developer's version of that breakdown: what each model really costs to operate, where "bring your own key" genuinely wins, and where a flat price earns its premium.
Wall one: no price on the page
A chunk of this category hides pricing behind "contact us." Legitimate for enterprise and usage-based contracts, but for anyone comparing options it means a sales call before you can even budget. Read "contact us" as: negotiated, probably higher, and a time cost to find out.
Wall two: "free" means "bring your own key"
This is the part worth understanding as an engineer, because it's the most common model and the one that looks cheapest on the surface.
A typical free AI search plugin is open on WordPress.org, but to make it work you connect your own OpenAI API key, and sometimes a Pinecone or Supabase instance on top. The plugin generates embeddings and runs queries against services you pay for directly.
Three things move onto your plate:
- A metered bill. Embeddings are genuinely cheap to index — for a small catalog you're realistically in single-digit dollars a month, and per-query cost is smaller still. The catch isn't the steady state, it's the variance: a full re-index, a traffic spike, or an LLM call in the query path changes the number, and it's hard to forecast in advance.
- The infra. If the plugin needs a vector database, you're now running a Pinecone or Supabase account, picking dimensions, and keeping it in sync. That's real setup, not a toggle.
- The operations. Keys rotate, accounts need a card on file (even for free credits), spending limits need watching. The AI backend is now something you maintain.
None of this is hidden maliciously. It's just a cost that lands after install, which is exactly why "free" is a confusing label here.
The honest part: when bring-your-own-key wins
If you're a developer, BYO-key is often the right call, and I'll say it plainly: on raw cost, you can beat any subscription. Small catalog, modest traffic, comfortable with an API key and a usage dashboard — paying OpenAI directly can run you a few dollars a month. You trade a predictable subscription for metered cost and full control, and for a technical owner that's a fair trade.
The trade-off is predictability, effort, and support. You own the setup, you carry the surprise-bill risk, and there's no managed support line at 2am. For a non-technical store owner who wants search to just work, that math flips.
The models, side by side
As of June 2026 (check each vendor's page, these move):
| Model | Published price? | Also on you |
|---|---|---|
| Free plugin + your OpenAI key (e.g. Mori) | No | Metered OpenAI bill; account + key |
| Free plugin + OpenAI + vector DB (e.g. OC3) | No | Two metered bills; self-hosted setup |
| Keyword by default, semantic via your backend (e.g. AI Vector Search) | Managed tier unpriced | Supabase + OpenAI for the real semantic mode |
| Freemium SaaS, key for real use (AI Search / wp-search.ai) | Free tier yes, premium not listed | Own key once you outgrow the free service |
| Published hosted SaaS (Motive) | From ~EUR 39/mo | Nothing extra, all-in |
| Hosting-locked (WP Engine Smart Search) | From ~USD 140/mo | Requires WP Engine hosting |
| Queryra | 14-day trial, then from USD 9.99/mo | Nothing — no key, no per-search bill |
Pattern: nearly every "free" true-semantic plugin runs on BYO-key, so its real cost is a metered bill you can't see up front. The flat-price products tend to be the more mature hosted services.
What you're actually buying
"Semantic" means matching on meaning, not exact words. Each item becomes an embedding — a numeric fingerprint of meaning — so "warm waterproof jacket under 100 dollars" can find the right coats even when those words aren't in the titles, and can read the price as a filter. Generating those embeddings and running queries through a model is the part that costs money. Whether that cost reaches you as a metered bill or a flat subscription is the entire pricing question. The search quality can be similar; the billing model is what differs.
Test before you pay
One practical filter that cuts through all of this: can you try the search on real data before you commit? Most plugins gate it behind install, signup, or a sales call. A public demo is a vendor saying "judge the quality yourself, first." Queryra keeps one open — search a live WooCommerce store, no signup. Whatever you pick, prefer the option you can actually test over the one you can only read about.
Where Queryra fits
Disclosure: I build Queryra, so treat this as the contrast, not a neutral verdict.
Queryra runs vector embeddings plus LLM intent parsing, hosted — no OpenAI account, no key, no per-search bill. The AI stack runs on our side; you install, connect once, sync. Pricing is flat and published: a 14-day trial, Starter from $9.99/mo, Pro at $199/mo, Enterprise custom. Current breakdown always on the pricing page.
To be clear: Queryra is not the cheapest. A dev on their own key beats any subscription on raw cost. What it offers is one predictable price with zero AI setup and real support. The price on the page is the price.
The checklist
- Is the price published, or "contact us"?
- Does it need my own OpenAI / Pinecone / Supabase key?
- Flat billing or usage-based?
- Actually semantic, or keyword with an AI label?
- Tied to a specific host?
- Can I test it on real data without a sales call?
Answer those six and the real cost stops being a surprise.
