# Pagefind Review: A Hands-on Test With Twelve Static Pages

> A reproducible functional test of Pagefind 1.6.0 using German and English HTML fixtures.

- URL: https://bytedossier.com/en/review/pagefind-review/
- Author: [Byte Dossier Editorial Team](https://bytedossier.com/en/author/editorial-team/), Editorial team
- Published: 13 November 2026
- Other languages: DE https://bytedossier.com/test/pagefind-test/

## In short

Pagefind review: Pagefind 1.6.0 passed ten checks covering search results, language separation, filters and excluded content in our small fixture. Twelve HTML pages were indexed. We executed the generated JavaScript/WASM search API under Node.js, not a browser interface. The findings support further evaluation for static content sites, but do not establish production performance.

## Verdict

**Rating: 8/10**

8.0/10 for the tested static-content use case. Validate actual content and browser behaviour before production.

**Pros**

- Ten predefined functional checks passed
- German and English separated in the fixture
- No additional search server installed for the test

**Cons**

- Index freshness depends on the build
- Browser operation and large corpora not evaluated

- All ten predefined functional checks passed.
- The fixture contained twelve pages across two language indexes.
- The score applies to this small test and use case.

## What did this Pagefind review actually test?

This Pagefind review used version 1.6.0 with twelve short HTML fixtures: six German and six English pages. Topics included GEO, WebP, AVIF and search. Each page had a data-pagefind-body main region, a topic filter and deliberately excluded text. Two additional landing pages lacked the main-region marker and were not indexed.

Pagefind generated two language indexes containing 111 indexed words. We then executed the generated JavaScript and WebAssembly search API in Node.js. A minimal document shim supplied language attributes, while filesystem-backed fetch replaced network requests. This was an executed functional test on a small artificial corpus. It did not test a browser interface or production website.

## Which checks passed?

Each case specified expected URLs in advance. German searches for “WebP” and “Pagefind” had one expected destination each. GEO was queried with matching and non-matching topic filters. Navigation-only text, a data-pagefind-ignore region and an unknown token had to produce no results. English checks covered two individual searches and one topic-filtered query.

| Check group | Cases | Result |
| --- | --- | --- |
| German individual queries | 2 | 2 passed |
| German topic filters | 2 | 2 passed |
| Exclusions and unknown token | 3 | 3 passed |
| English queries and filter | 3 | 3 passed |
| Total | 10 | 10 passed |

![Pagefind review: Pagefind functional test results: ten passed checks across twelve German and English fixture pages.](https://bytedossier.com/media/01a1d52d-22ee-7451-94cc-5c12bbe4b935-800.webp)
*Original Byte Dossier graphic. Own test of 13 November 2026; setup and limitations explained in the text.*

These results confirm expected matches and exclusions for this fixture. They do not establish ranking quality for ambiguous queries, typo tolerance or complex specialist terminology. Those questions need a larger, editorially evaluated query set.

## How large was the generated output?

The full generated directory contained 701,274 bytes across 31 files. This is not the transfer size of an individual search. Three separately timed CLI invocations, including process startup, took 135.74, 91.11, 126.79 milliseconds, with a median of 126.79 milliseconds. These observations come from one Linux environment and are not a hardware benchmark.

Actual transferred bytes, cache behaviour and responsiveness on weaker devices still need measurement. Keyboard operation and screen reader behaviour were also outside this test. A small build duration alone cannot demonstrate a good search experience.

## When is Pagefind worth evaluating?

Consider it when HTML already exists at build time and search may reflect that build's content. Documentation sites and editorial archives fit that model. If a site loads important text from a database at runtime, establish whether that text is present in the indexed build.

A catalogue with rapidly changing inventory needs a different freshness analysis. Do not promise real-time availability from an index refreshed only at build time. This test also did not evaluate semantic answer generation: successful keyword matching is not evidence of AI understanding.

## Why is the score 8.0 out of 10?

The editorial score is limited to this use case: basic search 9, language separation 9, tested filtering and exclusions 8, and integration requirements within the build model 6. Their unweighted mean is 8.0. The integration score reflects the required build step and additional browser validation; it is not a measured latency score.

Before deployment, generate the index after the site build, publish both together, verify language attributes on actual content and test real queries with expected results. Check browser accessibility and transfer costs separately. Use reliable rebuilds to remove outdated material from the index.

Cover: AI-generated thematic illustration, not an actual photograph.

## Which articles help with the next decision?

- [WebP vs. AVIF vs. JPEG](https://bytedossier.com/en/compare/webp-vs-avif-vs-jpeg-comparison/)
- [Pillow review](https://bytedossier.com/en/review/pillow-review-webp/)
- [All articles in this topic](https://bytedossier.com/en/topic/web-technology-performance/)

## Frequently asked questions

### Was this an executed hands-on test?

Yes. Indexing and the generated search API were executed on twelve artificial HTML pages. Browser UI and production load were outside the scope.

### Does every search download 701,274 bytes?

No. That was the size of the full generated directory. Per-query network transfer was not measured.

### Did you test semantic AI search?

No. The cases covered keyword matching, filters and language separation, not generated answers.

## Sources

1. [Pagefind 1.6.0 documentation](https://pagefind.app/docs/), Pagefind
2. [Multilingual search](https://pagefind.app/docs/multilingual/), Pagefind
3. [Using the search API](https://pagefind.app/docs/api/), Pagefind
4. Byte Dossier Lab: eigener Funktionstest vom 13.11.2026; 12 HTML-Seiten, 10 dokumentierte Prüfungen