Search function completely useless? Synonyms retired?

[gracemay]

Hi all. Anybody have any solutions or insights as to why Search & Discovery is suddenly utterly useless? When searching keywords on my store, the query is returning hundreds of products - almost all of which are completely unrelated to the search term.

When I visited the Search & Discovery App to try to adjust settings, I see that the Synonyms feature has been “retired,” meaning I can’t even adjust related keywords anymore.

I know Synonyms was active within the last few months, so I feel some major change has taken place recently. I don’t think its an understatement to say that this is REALLY BAD for business. If customers cannot adequately search for products, they are very, very likely to abandon the purchase.

PLEASE HELP.

[Ecom_swift_LLC]

Hello @gracemay, you’re not missing a setting, other merchants only started reporting Synonyms disappearing around Sep 16 with nothing in the changelog explaining it, so this looks like a platform-side change rather than something on your end. For the relevancy problem specifically, that’s usually separate from Synonyms, go to the Search & Discovery app > Search > Product boosts and pin your top sellers to the search terms customers actually use, that gives you a working stopgap while default ranking is off. I’d also open a ticket with Shopify Support directly since it’s hitting multiple stores, that gets it on their radar faster than the community can fix it.

If this helped, please mark it a solution

[Jack_BuildsShopify]

@gracemay Since this changed recently, I’d separate “Shopify changed ranking behavior” from “catalog fields are now matching too broadly.” Pick 10 real customer queries and make a tiny before/after relevance sheet: expected top products, actual top 10 results, and which fields on the bad matches contain the term (title, tags, product type, vendor, description, metafields). Product boosts can protect the highest-value searches while Support investigates, but that sheet will tell you whether the bad results share one field you can clean up instead of guessing. Roughly how large is your catalog, and are the unrelated matches sharing a common tag/product type or only words buried in descriptions?

[Ploqo]

Hi @gracemay

i built and tested on Dawn theme

Before (your problem)

After the fix

Steps

1. Go to Online Store.
2. Click Edit theme on your current theme.

3. Click Home page at the top and choose Search.

4. Click Add section on the left and pick Custom Liquid.

5. Paste this code in the Liquid box:

<style>.strict-search-wait .template-search{visibility:hidden}</style>
<script>
  (function () {
    var FIELDS = ['title', 'tag', 'product_type', 'variants.sku'];
    var params = new URLSearchParams(location.search);
    var q = (params.get('q') || '').trim();
    if (!q || params.has('loose')) return;

    var words = [], m, re = /\(title:(\S+?) OR /g;
    while ((m = re.exec(q))) words.push(m[1]);

    if (!words.length) {
      if (/[:"()*]|\b(AND|OR|NOT)\b/.test(q)) return;
      var strict = q.toLowerCase().split(/\s+/)
        .map(function (w) { return w.replace(/^-+/, ''); })
        .filter(Boolean)
        .map(function (w) {
          return '(' + FIELDS.map(function (f) { return f + ':' + w; }).join(' OR ') + ')';
        }).join(' AND ');
      if (!strict) return;
      document.documentElement.classList.add('strict-search-wait');
      params.set('q', strict);
      location.replace(location.pathname + '?' + params);
      return;
    }

    var original = words.join(' ');
    {%- if search.results_count == 0 %}
    document.documentElement.classList.add('strict-search-wait');
    params.set('q', original);
    params.set('loose', '1');
    location.replace(location.pathname + '?' + params);
    return;
    {%- endif %}

    function tidy() {
      document.querySelectorAll('input[name="q"]:not([type="hidden"])').forEach(function (i) {
        if (i.value === q) i.value = original;
      });
      var walk = document.createTreeWalker(document.body, NodeFilter.SHOW_TEXT);
      while (walk.nextNode()) {
        var n = walk.currentNode;
        if (n.nodeValue.indexOf(q) > -1) n.nodeValue = n.nodeValue.split(q).join(original);
      }
      if (document.title.indexOf(q) > -1) document.title = document.title.split(q).join(original);
    }
    document.addEventListener('DOMContentLoaded', function () {
      tidy();
      new MutationObserver(tidy).observe(document.body, { childList: true, subtree: true });
    });
  })();
</script>

6. Click Save.

Search now shows only products with the words in the title, tags, product type or SKU.
If nothing matches (for example a typo), it shows the normal results.

Regards,
Ploqo

[alaattincagil]

Hi @gracemay , this isn’t a bug on your end, it’s the same change hitting a few other threads right now (there’s a “Search results” thread from a few days ago with the exact same “hundreds of unrelated products” symptom, duck earrings pulling in moons and stars). Shopify rolled out semantic search on Search & Discovery, and Synonyms getting retired is downstream of that. The whole point of semantic matching is that it looks at description text, image data, categories and related concepts instead of you manually mapping “sneakers” to “athletic shoes”, so the feature that used to do that by hand got folded into the automatic one. It’s in Shopify’s help center under Search & Discovery, semantic understanding, just not linked anywhere you’d find it when Synonyms disappears from the app.

Two things worth trying while you wait on Support (still worth opening a ticket, Support confirmed to at least one other merchant this is intentional and not something you can toggle off):

Product boosts in Search & Discovery won’t shrink the result set, but they’ll push your actual matches to the top, which is most of what a customer sees before scrolling past.

For any specific query you want strict matching back on, title:word1 AND title:word2 turns semantic matching off just for that search. Stricter than Synonyms was (no fuzzy matching at all), but it kills the 900-unrelated-results problem instantly if you’re testing a term.

Before spending time rewriting descriptions to influence the semantic side, I’d pull the “Searches with no clicks” report in Search & Discovery first. That separates queries actually costing you sales from ones that just look messy in a manual test. We run that triage with clients before touching catalog data, since irrelevant results and zero results usually need different fixes, and it’s easy to burn a week cleaning descriptions for a term nobody was typing.

Hope this untangles it a bit, will drop back if anything changes on the rollout.

[IFGeCommerce]

My name is Francesco Guiducci and I’m the founder of IFG eCommerce.

I would avoid adding a theme script that rewrites every customer search into field-specific search syntax as the first fix here.

Shopify’s current documentation is important in this case: there is no setting that controls how broadly native storefront search matches, and ranking behavior can also be tested and changed on Shopify’s side. The current Search & Discovery documentation no longer presents custom synonyms as the control merchants previously had, so I don’t think you’re simply overlooking an old setting.

I would first confirm that the store is actually using Shopify’s native search. A third-party search app, a custom Storefront API implementation or theme-level search customization can produce completely different results.

If it is native search, take a small set of real customer queries and record the expected products versus what Shopify currently returns. For the irrelevant products, check whether the search term appears in the title, description, product type, vendor, tags, SKU or barcode. That gives you something reproducible rather than trying to tune the catalogue blindly.

Product boosts can help important products rank higher for selected searches, but they don’t solve the underlying problem if hundreds of irrelevant products are matching in the first place.

I would also be cautious with the Custom Liquid workaround posted above. Shopify documents that using search syntax changes native search behavior and disables features including typo tolerance, predictive search and semantic understanding. Applying that transformation globally could therefore fix one query while creating different search problems elsewhere.

If the same query repeatedly produces clearly unrelated results using native search, I would send Shopify Support the exact query, the product you expected, examples of the incorrect results and the approximate test time. Those are the details Shopify currently asks for when investigating search-result problems.

Until the cause is clear, I would diagnose the matching rather than replacing Shopify’s search behavior at theme level.

[VikashJ]

Hi @gracemay,

This isn’t something specific to your store. There’s been a broader, unannounced rollout to Shopify’s search engine this year, and a number of merchants across different stores and languages have reported the same pattern: products that used to be findable are suddenly no longer findable, and relevance has gotten noticeably worse. Synonyms being retired without warning fits that same pattern of a platform-side change rather than anything you did on your end.

While you wait for it to stabilize or for Shopify to respond, here’s what’s worth doing. It helps to know that Shopify’s predictive search, the dropdown that appears as someone types, and the full storefront search results page after they hit Enter, actually run on two different engines, so a fix or workaround for one won’t necessarily carry over to the other. Also worth knowing: predictive search doesn’t search the SKU field by default, it only searches title, product type, variant title, and vendor, so if the irrelevant results you’re seeing involve SKU-like or code-style search terms, that’s a separate, longstanding limitation rather than a new bug.

For an immediate stopgap while relevancy is broken, Product boosts in Search & Discovery let you manually pin your correct products to the search terms customers actually type. That won’t fix the underlying relevancy engine, but it does correct the specific searches customers use most. If you were relying heavily on synonym groups for a catalog with lots of alternate naming, like color names, regional spelling, or abbreviations, consider temporarily adding those alternate terms directly into product titles, tags, or the product type field, since those are the fields native search still reads reliably.

If your catalog is large, also check whether the issue is worse on bigger collections specifically, since native filters have been reported to become unreliable once a collection crosses several thousand products, which could be compounding what you’re seeing.

Since this is affecting multiple stores at once rather than just yours, filing a support ticket still helps get it tracked on Shopify’s end, and searching the community for other recent threads reporting the same synonym removal is worth doing too, since a visible pattern across many merchants tends to get attention faster than one ticket alone.

Best,
Vikash Jha - Apploy

[devcoders]

Hi @gracemay

Welcome to the Shopify Community! Please share your store URL and password (if it’s password-protected), so I can check and provide you with the exact solution.

Best regards,
Devcoder

[adriclico]

I’m actually working on this exact relevance problem with Vectail, an ecommerce search engine built on Google’s commerce search technology.

The goal is to keep semantic understanding without letting loosely related products flood the results like this.

It’s free to try if you want to compare the same searches on your catalog.

[codeofsagar]

Hi @gracemay, you’re not alone, it’s the same change as the “duck earrings” thread.

On Ploqo’s snippet and Francesco’s warning: both are right. Forcing searches into title/tag matching brings precise results back, but those searches lose typo tolerance. The safe middle ground is to keep the strict version only when it finds something and fall back to normal search when it finds nothing. Ploqo’s code already does that fallback, so if you try it, test a few misspelled searches before leaving it on.

For the synonyms you lost: if you still have your old list, even from memory, adding the alternate words as product tags is the closest free replacement.

Disclosure: I’m a developer building a small app for exactly this. It keeps your normal search page and filters, brings back a synonym editor, and adds exact matching with that fallback built in. It’s not in the App Store yet; I’m looking for a few stores to test it free (paid plans will be $9–19/month after launch). If you’d like to try it, say so here.

[Cotelo]

Hi @gracemay,

I’m very curious to learning more about your pain points in regards to the Search and Discovery problem you’re facing and how you’re currently approaching to solve this problem.

Please feel free to reply back to me here as I would love to help generate ideas to help you come up with a solution that solves your problem.

On a side note, I’m wondering how many products do you currently have on your shopify store?

[DanielAnderson]

You’re not missing a setting. It looks like Shopify switched native search to semantic matching, which reads descriptions and categories rather than just the words typed, and that’s why a search now pulls in anything loosely related. Retiring Synonyms seems to be part of the same change.

Before trying any fix, pull your top search terms from the Analytics search report for the last 30 days. The top 20 are probably most of your search revenue. Run each one yourself and boost the right products on any that come back wrong. That’ll take a bit but it will cover the searches that actually matter while Shopify sorts itself out.

Separately, I run Carti, an AI sales assistant for Shopify, and this is close to the exact problem it’s built for. It doesn’t replace your search bar, but a shopper who asks for something like ‘a gift for a dad who fishes’ will get three strong product recommendations rather than a wall of 200 random options. If you’re interested, the app is totally free at your size, only about five minutes to set up, and it works alongside whatever you choose for native search.