What BigCommerce Conversational Search actually changes
BigCommerce's developer changelog entry dated October 29, 2026 announces Conversational Search, an upgrade to the platform's native storefront search. The entry says shoppers can search in natural language and still find relevant products, even when their words don't appear in the merchant's product data. BigCommerce calls the mechanism hybrid search: meaning-based matching combined with keyword matching. It runs on embeddings that BigCommerce generates for the catalog in the background.
In practice, this changes how products match descriptive queries in stores that turn the feature on. A product's appearance for a query phrased around a use case or an attribute now depends partly on how BigCommerce's embedding model represents that product. Word overlap with the title or description still counts, but it is no longer the only factor. Per the changelog, SKUs, product names and brands keep working. Merchants can opt in or out at any time under Settings › Search, and the feature costs nothing extra on paid plans. This is a change to on-site search only. The changelog makes no claim that it affects how ChatGPT, Perplexity, Google AI Overviews or other external engines find and cite products.
Keyword search versus hybrid search, side by side
The changelog doesn't describe the previous algorithm in detail. It presents the change as a fix for searches where the shopper's wording doesn't match the product data. That implies the old native search relied mostly on matching the shopper's words against the catalog. The table separates what BigCommerce states from what it leaves unstated.
| Aspect | Before (implied by the changelog) | With Conversational Search (per the changelog) |
|---|---|---|
| Matching method | Shopper words matched against product data | Meaning-based matching combined with keyword matching |
| SKU, product name and brand queries | Matched by keyword | Stated to “keep working” |
| Descriptive, natural-language queries | Could return nothing when wording differed from product data | Stated to return relevant results |
| Catalog preparation | Not applicable | Embeddings generated in the background; no copy rewrite required |
| Zero-result searches | Baseline | “Up to 70% fewer,” per BigCommerce |
| Cost and control | Not applicable | No additional cost on paid plans; opt in or out under Settings › Search |
| Release stage | Not applicable | Not stated |
What it means for BigCommerce brands
Our read is that the change matters most in the long tail of on-site search. Some shoppers type a product name or a SKU code such as MERINO-BL-MD. Those navigational queries should behave as before, because BigCommerce keeps keyword matching for exactly those cases. The queries that change are the ones phrased around a use case, an attribute or a problem. That is the kind of phrasing shoppers bring over from AI assistants.
Three consequences follow. All three are our inference, not something BigCommerce states:
- Part of retrieval control moves to the vendor. For descriptive queries, results depend partly on BigCommerce's embedding model. The changelog gives merchants no stated control over that model, its inputs, or how it is weighted against keyword matches.
- “No catalog changes” does not mean copy stops mattering. The embeddings are presumably generated from catalog data the merchant already maintains. A vague title or thin description gives any model less to work with. BigCommerce hasn't said which fields are embedded.
- Zero-result pages become a weaker signal. Empty results used to reveal missing products or vocabulary the catalog didn't use. If meaning-based matching fills those gaps, merchants lose that rough but useful signal. Search reviews will have to look for irrelevant results, not just empty ones.
Teams already rewriting titles for AI search should keep going. Specific, attribute-rich titles, of the kind covered in how to rewrite product titles for AI search, give both keyword and meaning-based matching more to work with.
Reading the 70% claim and the no-rewrite promise
BigCommerce's changelog states that stores see “up to 70% fewer searches that return no results.” That is an upper bound. The entry gives no methodology, no sample of stores, no time frame and no definition of the queries compared. Our read is that it describes BigCommerce's best observed case, not a typical result. How much a given store gains will depend on how often its shoppers' wording differed from its catalog vocabulary to begin with.
The same caution applies to “No catalog changes.” BigCommerce's claim is that merchants don't need to rewrite product copy for the feature to work. That is plausible, since BigCommerce builds the embeddings in the background. Whether existing copy is good enough to match descriptive queries well is a separate question, and the changelog doesn't address it.
Fewer empty results pages measures recall. It says nothing about whether the products returned are the right ones.
Why this will not move your AI engine citations
Conversational Search runs inside the BigCommerce storefront, and BigCommerce's changelog makes no claim that it affects external AI engines. Our understanding is that ChatGPT, Perplexity, Google AI Overviews and AI Mode, Gemini, Claude and Copilot build product answers from crawled pages, structured data and their own retrieval systems. They don't use a store's internal search index. Turning the feature on should not be expected to change how often those engines cite a store's products.
For external engines, the levers are still crawlable PDP content, complete Product structured data, and letting AI crawlers in. How AI engines pick which products to cite and schema for AI search cover those mechanisms. There is still a real link between the two. BigCommerce is adapting on-site search to the same natural-language phrasing that AI engines handle, and specific, accurate product data helps in both places.
What BigCommerce has not said yet
The changelog entry links to a separate Conversational Search documentation page. That page was not part of the source material for this analysis. Until merchants read it, these questions remain open:
- Default state. Whether the feature is on or off by default for eligible stores.
- Plan coverage. Which plans count as “paid plans,” and whether availability is the same across tiers.
- Storefront coverage. Whether it covers only Stencil storefront search, or also Catalyst and headless builds that use the GraphQL Storefront API's
site.search.searchProducts. - Embedded fields. Which catalog fields feed the embeddings: name, description, custom fields, variant options, categories, brand or SEO fields.
- Model and refresh rate. Which model generates the embeddings, and how quickly they update after a product is created or edited.
- Languages and locales. Which ones are supported.
- Rollout and release stage. The entry has no beta, preview or GA label and doesn't describe a staged rollout.
- Tuning. Whether merchants can adjust the balance between meaning-based and keyword matching, or apply merchandising rules and boosts on top.
- Shopping Agent. Whether Conversational Search supplies product retrieval for BigCommerce's Shopping Agent.
Headless teams should also know about a related fix. A separate October 29, 2026 changelog entry states that BigCommerce fixed a Storefront GraphQL API bug. In that bug, site.search.searchProducts could leave out products that matched every filter in filters.productAttributes. The entry says the fix is live for all stores. The changelog does not connect it to Conversational Search. Developers checking their integration can start with BigCommerce's API overview and quick-start documentation. How to audit a headless storefront for AI readiness covers the external side.
What to do this week
- Read the linked documentation. Before changing anything, use the BigCommerce developer changelog entry to reach the Conversational Search docs. Confirm plan eligibility, the default state and which storefront types are covered.
- Check Settings › Search. Decide on purpose whether to opt in or out. If the feature is already on, find out when it was turned on so your before-and-after comparisons have a start date.
- Export a baseline. Pull your current on-site search analytics: zero-result queries and the top descriptive queries. Without a before snapshot, you can't compare your store against BigCommerce's “up to 70%” figure.
- Run a fixed set of test queries. Write 20 to 30 representative queries, split between navigational (SKU codes, product names, brands) and descriptive (use-case, attribute or problem phrasing). Run the same set before and after. Record whether the top results are relevant, not just whether any results appear.
- Tighten the catalog data you control. Keep titles, descriptions, attributes and custom fields specific and accurate. No rewrite is required for the feature to run, but the embeddings come from this data.
- Keep external AI work on its own track. Conversational Search doesn't replace crawlable PDPs, Product structured data or AI crawler access. Start with where AI engines get information about your products.
The strategic read
The same changelog page has an October 30, 2026 entry about BigCommerce's Shopping Agent. It says BigCommerce is starting to add a BETA tag to the feature in the control panel to show it is in open beta. Our interpretation is that, taken together, the two entries show BigCommerce investing in conversational, agent-style discovery on the storefront itself. The changelog doesn't connect the two features, and Conversational Search may simply be a search-quality upgrade that stands on its own.
Either way, our read is that the direction matches where shopper behavior is heading. Queries look more like questions to an AI assistant, and platforms are moving retrieval from keyword overlap toward meaning. For merchants, the strategic consequence is that catalog data now feeds two kinds of machine reader: the platform's own embedding model and external AI engines. The merchant directly controls neither. The best response is the same for both: complete, specific, accurate product data, measured before and after every platform change instead of judged against a vendor's upper-bound figure. Brands that want to track that systematically can see how product-level AI visibility for D2C brands works.
Frequently asked questions
What is BigCommerce Conversational Search?
According to BigCommerce's developer changelog entry of October 29, 2026, it is an upgrade to native storefront search. It combines meaning-based matching with keyword matching so shoppers can search in natural language and still find relevant products, even when their words don't appear in the product data.
Do I need to rewrite my product copy to use Conversational Search?
BigCommerce's changelog states that no catalog changes are needed, because it generates embeddings for the catalog in the background. The embeddings presumably come from your existing catalog data, though, so specific and accurate titles, descriptions and attributes likely still affect how well products match descriptive queries. BigCommerce hasn't said which fields are embedded.
Will SKU and brand searches still work with hybrid search?
Yes, according to the changelog. It states that under hybrid search, SKUs, product names and brands keep working while descriptive queries return relevant results.
Does Conversational Search improve visibility in ChatGPT or Google AI Overviews?
BigCommerce makes no such claim. Conversational Search changes on-site storefront search. External AI engines rely on crawlable PDP content, Product structured data and crawler access, so those remain the levers for AI citations.
How much does BigCommerce Conversational Search cost and how do I turn it on?
The changelog states it is available at no additional cost on paid plans, and merchants can opt in or out at any time under Settings › Search. It doesn't say which plans qualify or whether the feature is on by default.
Ask AI about this
Ask an AI engine how BigCommerce's hybrid search differs from keyword search and what it means for your store's catalog data.