This page is in three parts: what all three of us do and how ours differs, what only ours does, and what only theirs does. The third part is the same size as the other two, because a comparison that only lists its own wins is an advertisement. Every row names its source and its date.
Edge-native search and merchandising engine. Runs 100% of search on one live 24,875-product store since Aug 24, 2026, and on a public demo. Not in the Shopify App Store.
The Shopify category leader. Listed since Jan 2017, 4.8★ across ~1,500 reviews, Built for Shopify badge. Priced in GMV bands from $29.
Enterprise search platform, ~$230M ARR in 2025 (Sacra estimate, third-party). Metered per search; AI features largely behind enterprise contracts.
SearchSmith is an edge-native search and merchandising engine for Shopify-style catalogues. As of August 2026 it runs 100% of search and browse on one live 24,875-product store and on a public demo, and it is not listed in the Shopify App Store.
Boost AI Search & Filter is the Shopify category leader, listed since January 2017, rated 4.8 stars across roughly 1,500 App Store reviews, carrying a Built for Shopify badge, and priced in GMV bands. Algolia is an enterprise search platform at roughly $230M ARR in 2025 by a third-party estimate, whose AI search and personalization sit largely behind enterprise contracts.
Measured against those two, SearchSmith ships five capabilities neither competitor offers and behaves differently on ten more that all three provide. It is behind on twelve, most importantly App Store availability, ecosystem integrations, multi-language support, back-in-stock email delivery and a published uptime commitment.
Ten capabilities every one of these products ships. The difference is never the checkbox — it is the mechanism underneath, and each one below names it.
Boost ships semantic search on its $29 Launch tier as an impact-weight slider, default 50%, with no visibility into how lexical and semantic results were blended. Algolia sells it as NeuralSearch, on the Elevate annual enterprise contract only.
Every query runs embeddings, tiered full-text and SKU/barcode exact-pins, fused by reciprocal rank fusion with similarity floors that reject gibberish — on 100% of storefront queries, at no tier, with the whole blend readable per result.
All three ship pin, boost, bury, hide and scheduling. Boost added margin merchandising in Feb 2026: manual threshold rules on margin percent, using Shopify's cost-per-item. Algolia lets you wire margin in as a custom-ranking attribute.
The same rules, plus one neither has: when two products fit the query about equally well, the one with more gross profit in dollars goes first — automatically, only inside 25% of the top relevance score, never over an exact SKU match or a pinned product, with a per-row audit line and a live kill-switch.
All three ship search analytics. Boost documents manual IP exclusion. Algolia caps per-user event counting to stop gaming, but its docs describe no bot classification of analytics.
On the production store, 73% of sessions are bots. Every merchant-facing number rides a humans-only filter — user-agent, webdriver signal, velocity flag — labelled as such throughout the console, with the bot share available in the analytics tab and no raw IP or user-agent stored. If your current app does not do this, the search volume and click rate you have been reporting include the robots, and you have no way to know by how much.
Both can rank on availability, and both need the app installed with API access to get the data.
A crawl every 20 minutes with no store credentials drives a sink on sold-out products, in-stock-first partitioning on page one, in-stock-only suggestions and fallbacks, and a coverage guard that fires below 95%. Availability is not just a filter: sold-out@10 is one of the metrics the deploy gate scores.
Boost's no-results page shows popular terms and recommendations. Algolia offers removeWordsIfNoResults parameters you configure.
Four stages — rerun the corrected spelling, relax one term at a time, an in-stock grid sorted by profit, then clear-filters — and the search is still logged as a zero-result, truthfully, feeding a queue of rule candidates you can act on. A dead end becomes a captured customer and an honest number, not a hidden one.
Algolia exposes ranking info through its API and dashboard. Boost gives you a preview tool.
Every staff row carries an audit of which rules fired, the profit factor, the numeric factors and the popularity prior — with an independent kill-switch per factor that works on live production traffic, one query parameter each.
Boost's widget on this store, build 1.88.2, measured on a collection page before it was removed: 102,571 bytes gzipped of first-paint JavaScript and CSS. Boost shipped TURBO — a server-rendered theme app extension — in March 2026, and we have not measured that build.
7,509 bytes gzipped on first paint — the loader that runs on a page where search is just the box in your header — measured with the same compression on the same store. Fully warmed, with every lazy chunk loaded for a full results page, the whole widget is 74,470 bytes: still less than the figure above. The byte budget is step zero of every deploy; the build fails if the loader grows.
Boost gates in-collection search behind its Convert tier — listed at $299–$599/mo on the Shopify App Store and $239/mo on boostcommerce.net. Those are two of Boost's own price sheets, and they disagree; both fetched Aug 23 2026.
On the default collection template of every collection, with Load more and a server-rendered fallback, at no tier cost, in every deployment.
Boost ships merchandising A/B testing. Algolia ships codeless A/B with a sample-size estimator and a frequentist two-tailed test.
The same two-proportion z-test with a p-value, a significance flag and a needed-sample estimate — and it refuses to show a verdict under 30 searches per arm. The difference is where the restraint sits: Algolia's estimator tells you how much traffic you would need and leaves the reading to you; SearchSmith withholds the verdict until the traffic is there.
Algolia builds a Query Suggestions index daily from 30 days of analytics. Boost returns suggestions from a dictionary the merchant curates.
Completions built from the store's own human searches with no curation and no seeding, plus did-you-mean correction and collection and guide chips. And when nothing matches exactly the widget says so — “no exact matches, here are the closest” — instead of showing loose matches as though they were hits.
Five capabilities. Each states what was searched to justify the absence, because an absence claim without a stated search is just an assertion.
You have had an app update make your store worse. Every production release is scored on a preview URL against a 150-query evaluation set with 3,089 graded relevance judgments before it is promoted, and promotion is blocked automatically when relevance drops past a published threshold — nDCG@10 by more than 0.02, profit-weighted nDCG by more than 0.03, sold-out @10 rising by more than 0.01, or the zero-result rate rising by more than 0.02. Three ranking ideas were built, deployed to a preview and measured before being switched off: a cross-encoder re-ranker and an LLM query-expansion lane, which made natural-language queries 0.027 worse, and a body-text re-ranking pass.
Neither competitor gates merchant-visible ranking changes on measured quality; their simulators and A/B tools run after the change is live. Searched across six independent sweeps of both vendors' documentation, changelogs and listings, Aug 22–27 2026.
Boost has no conversational assistant anywhere — not on its site, its changelog or its listing. Algolia's is contact-sales, and you pay your own LLM bill on top.
Ours has been public on the search page since launch, free, answering from the store's own guides and policy pages. Every sentence carries a numbered citation that is machine-checked, and anything that cannot be traced to a source is deleted before the shopper sees it. If no guide covers the question, it says so instead of guessing.
What that proves is traceability to a source, not that the source is right.
Point a phone at a barcode and it pins the exact product — native browser barcode detection straight into the SKU and EAN exact-match lane. No frame ever leaves the device.
No barcode capability appears anywhere in either competitor's inventory: six independent inventory sweeps across both vendors, Aug 22–27 2026, cataloguing 258 features in total.
No admin token, no API key, no permission to read your orders. Stock, prices and collections come from crawls of pages that are already public, every 20 minutes, and your cost-per-item never leaves our servers. A compromise of us cannot leak an admin token we were never given.
Both competitors require the app installed with API access to do the same job.
Every capability on this page is on by default in every deployment. Boost gates in-collection search, collection merchandising and banners at Convert; real-time sync, unlimited filter trees and AI-personalized ranking at Accelerate; and 90, 180 or 365 days of analytics retention by plan. Algolia gates NeuralSearch and AI Collections at Elevate; Dynamic Re-Ranking and Advanced Personalization at Grow Plus; the Merchandising Studio and revenue analytics at Premium or Elevate.
There is no tier here to be on the wrong side of.
Twelve capabilities where a competitor is ahead. Same size as the other two sections, and each one says whether it is a gap we intend to close, a deliberate non-goal, or a strength that measures a different axis than search quality.
Both are one-click installs from the Shopify App Store with Shopify-native billing. We are not in the App Store. Today the engine installs by custom distribution, which Shopify limits to a single store or one Plus organisation and which cannot bill through Shopify, so billing is direct. A listed app is roughly six months of full-time build, or nine to ten months at three days a week. This is the single biggest reason to buy theirs instead, and no amount of feature advantage answers it if you need search working next week.
Boost has been listed since January 2017, holds 4.8 stars across roughly 1,500 reviews and carries a Built for Shopify badge. We have one production store, one public demo and the receipts on this page. Track record is real evidence about a vendor's survival and support, and it is evidence we do not have — but it is not evidence about relevance, which is what this page measures.
Reviews apps, Klaviyo, Attentive, Weglot, Shopney, GemPages and more. We have none of them. That surface area belongs to the app wrapper rather than the engine and is scoped inside the app build, but today it is simply a gap — if your stack depends on one of those integrations, we do not fit yet.
Boost runs the whole loop — the bell, subscriber management, restock emails — through Klaviyo today, with native delivery on its roadmap. We capture the demand, flip it to ready inside the stock crawl, price the queue in profit dollars and export it, then hand it to your email platform. Sending is a written non-goal rather than an oversight, and it is the most-requested thing we do not do.
Boost does six storefront languages and syncs Shopify Markets. We do English. If you sell in more than one language or market, this alone should decide it.
Algolia runs multi-region infrastructure with SLAs and a monitoring API. Boost publishes a status page showing 100.0% uptime over 90 days, with no incidents in the window we checked. We publish neither a status page nor an SLA, and we will not pretend to one until there is a contract behind it. If procurement needs a signed uptime commitment, that is Algolia's row.
Algolia ships Advanced Personalization and Dynamic Re-Ranking from per-user behaviour, and Boost ships AI-personalized ranking on Accelerate. We rank the same way for everyone, deliberately for now — but this is a genuine capability gap, not a philosophical stance, and it is on the roadmap.
Boost ships all of them as part of a broader merchandising suite. These are documented non-goals here: this is a search and merchandising engine, not a commerce-features suite, and pretending otherwise would make it worse at the thing it is for.
Algolia offers InstantSearch libraries, SDKs for essentially every language, and an integration marketplace, at roughly $230M ARR in 2025 by a third-party estimate. We offer one widget and a documented HTTP API. If you are building a custom front end across several platforms, that ecosystem is worth real money.
Both have staffed support teams; Boost's reviews frequently name individual support engineers. Support here is one person. That is a real difference in what happens at 2am on Black Friday, and it should weigh against us. What sits on the other side of it: the person who answers is the person who wrote the ranker, so there is no first-line script between you and a fix — and because the engine holds no credential of yours, turning it off is a theme setting you control, not a support ticket you wait on.
Algolia ships an image-embedding model that recommends visually similar products from product photos, and both vendors ship richer recommendation modules than ours. We have not built visual search at all. (Drag-and-drop merchandising and voice search we do ship — those are in the console and the search bar today.)
Boost and Algolia are funded businesses with teams, roadmaps and continuity plans. If the person who wrote SearchSmith stops, the story changes. That is the honest shape of the risk, and buying from an established vendor is a legitimate way to avoid it. What bounds the damage is that we hold nothing of yours: no admin token, no API key, no copy of your orders. Removing the engine is a theme setting, and your catalogue is untouched underneath it — the exit is cheap by construction, which is not true of an app that has been writing to your store.
Is your search a cost you are minimising, or a lever you are pulling? If it is a cost — you want it installed, working, and never thought about again — Boost is nine years better at that than we are, and this page has now said so twice. If it is a lever — you want to know what a ranking change did to margin, you want the numbers to exclude robots, and you want a release blocked when relevance drops — then the five capabilities in part two are not available anywhere else, and you can test every one of them in the console below without an account.
No. As of August 2026 SearchSmith is not listed. It installs today only by Shopify custom distribution, which is limited to a single store or one Plus organisation and cannot bill through Shopify. Boost AI Search & Filter has been listed since January 2017.
Track record, ecosystem, languages, published uptime and packaging. Boost has ~1,500 App Store reviews and a Built for Shopify badge, 30+ integrations, six storefront languages, Shopify Markets sync and back-in-stock email delivery. Algolia offers contractual SLAs, multi-region infrastructure and a monitoring API. SearchSmith does none of those.
Five things: a deploy-time relevance gate that blocks a release which ranks worse; free citation-verified answers drawn from the store's own guides; barcode scanning into exact match; operation with no store credentials at all; and no feature tiers.
Boost prices in GMV bands from $29, with two published sheets that disagree (both fetched Aug 23 2026). Algolia meters per search from $0.50 per 1,000. SearchSmith pricing is not published; the stated intent is to price by catalogue size rather than by merchant revenue.
The merchant console behind the demo store is open. No sign-in, no account, no email.
Pin a product, run the A/B test, read the analytics, misspell something.
Not ready to click around? Send your store URL and I will run your own catalogue and your last 90 days of search terms through it, and show you what changes.
How this page was built. The competitor columns come from the vendors' own documentation, Shopify App Store listings, pricing pages and changelogs, fetched Aug 22–27 2026 across six independent inventory sweeps of roughly 45 features each. Where a vendor publishes two conflicting numbers, both are cited. An absence means the capability does not appear in that vendor's public documentation — an absence in their docs, not proof they lack it. The SearchSmith column reflects the engine as of the releases of Aug 28–29, 2026. The 3,089 relevance judgments were seeded by model on Aug 22, 2026; human grades entered in the staff tool override seeds wherever they exist. Competitor pricing, review counts and revenue estimates are dated because they will drift. If you find a row that is wrong, tell us and it gets corrected.