“Did You Mean” Spelling Suggestions in Better Search Pro

What is “Did You Mean”?

When a visitor’s search returns zero results — often because of a typo or misspelling — “Did You Mean” suggests a corrected search term instead of leaving them with an empty results page. Depending on the mode you choose, Better Search Pro either shows a clickable suggestion link or transparently re-runs the search with the corrected term.

Why Use It?

  • Recovers dead-end searches: Turns a “no results found” page into a productive one.
  • Learns from your site: Suggestions are drawn first from your own search log, so corrections reflect what your visitors actually search for.
  • Falls back gracefully: If your search log has no close match, Better Search checks your published content, and optionally the server’s spellchecker.

This is especially useful for sites with long-tail search traffic, where visitors frequently misspell product names, brands, or uncommon terms.

How to Enable “Did You Mean”

Better Search Pro adds these settings:

  1. Go to Better Search → Settings in the WordPress admin.
  2. Open the Features tab and turn on “Did you mean” suggestions.
  3. Save your changes. The Search tab will show the remaining “Did you mean” settings.
  4. Open the Search tab and choose a “Did you mean” mode:
    • Suggest (“Did you mean”): Shows a “Did you mean” link on the zero-results page. The original (empty) results are still displayed.
    • Auto-correct: Transparently re-runs the search with the corrected term when it actually returns results, and shows a link back to the original query.
  5. Set Minimum searches to qualify as a suggestion — a term must have been searched at least this many times in your search log before it can be suggested as a correction. Default: 3.
  6. Optionally, enable Use enchant as a fallback to fall back to the server’s enchant spellchecker when neither your search log nor your site content has a close match. The setting shows the extension’s detected status and is disabled if enchant isn’t installed on your server — see Installing enchant on your server below.
  7. Save your changes.

How Suggestions Are Found

Better Search checks each unmatched search term against three corpora, in order, stopping at the first one that finds a match:

  1. Search log corpus: Terms your visitors have searched at least the number of times set in Minimum searches to qualify as a suggestion.
  2. Content-index corpus: Words drawn from published post titles and public taxonomy term names, kept in a dedicated dictionary table that refreshes automatically twice daily and whenever a post is saved.
  3. Enchant fallback: The server’s enchant spellchecker, if the setting is enabled and the extension is available.

Quoted phrases ("like this") and terms shorter than the Minimum characters setting (Search tab, default 4) are left alone — there’s too little signal in a short token to correct safely.

ⓘ “Did you mean” is automatically skipped for a search that already returned results — it only activates on zero-result searches.

Installing enchant on your server

The enchant fallback needs PHP’s enchant extension. Most managed WordPress hosts don’t enable it, but if you run your own dedicated server or VPS, you can install it yourself.

Check whether it’s already installed

The Use enchant as a fallback setting shows a live status message: green if the extension is detected, red if it isn’t. You can also check from the command line:

php -m | grep enchant

Install the extension

Enchant itself is a spellchecking library; PHP’s enchant extension is a wrapper around it, and it also needs at least one backend dictionary provider (aspell or hunspell) plus a language dictionary package to have anything to suggest from.

Ubuntu/Debian

sudo apt install php-enchant enchant-2 aspell aspell-en

Swap aspell-en for the package matching your site’s language (e.g. aspell-fr for French, hunspell-de-de for German).

RHEL/CentOS/AlmaLinux

sudo yum install php-enchant enchant2 aspell aspell-en

macOS (Herd / Homebrew PHP)

brew install enchant
pecl install enchant

Then add extension=enchant.so to your php.ini if the install doesn’t do it automatically.

After installing, restart PHP-FPM (or Apache/mod_php) so the extension loads:

sudo systemctl restart php8.3-fpm   # match your PHP version

Then reload the plugin’s Search settings tab — the status message should switch to “The enchant extension is installed on this server.”

Choosing a dictionary locale

Enchant looks up a dictionary matching the locale from the bsearch_did_you_mean_enchant_localeOpens in a new window filter (default en_US). If your site’s content is in another language, make sure the matching dictionary package is installed (e.g. aspell-de for German) and filter the locale to match:

add_filter( 'bsearch_did_you_mean_enchant_locale', fn() => 'de_DE' );

If no dictionary exists for the requested locale, the fallback silently returns no suggestion rather than erroring — the search-log and content-index corpora are unaffected either way.

Theme Integration

If Seamless Mode is off, Better Search’s own search template already displays “Did you mean” and auto-correct notices automatically — no extra work needed.

If Seamless Mode is on, your theme renders its own search.php, so call these template tags directly wherever you want the notice to appear:

the_bsearch_did_you_mean()

Displays a “Did you mean: <suggested term>?” link when Suggest mode found a correction but the original search still returned no results.

<?php the_bsearch_did_you_mean(); ?>

Accepts an optional $args array with before and after markup, defaulting to <p class="bsearch-did-you-mean"> / </p>. Returns an empty string when there’s no suggestion for the current query — safe to call unconditionally. A get_bsearch_did_you_mean()Opens in a new window variant returns the markup instead of echoing it.

the_bsearch_autocorrect_notice()

Displays a “Showing results for <corrected term> — search instead for <original term>” notice when Auto-correct mode transparently corrected the query.

<?php the_bsearch_autocorrect_notice(); ?>

Accepts the same before / after arguments, defaulting to <p class="bsearch-autocorrect-notice"> / </p>. A get_bsearch_autocorrect_notice()Opens in a new window variant returns the markup instead of echoing it.

Rebuilding the dictionary

The content dictionary is built from your published post titles. Rebuilds read titles in batches of 2,000 posts rather than loading every title at once, which keeps memory use flat on large sites. Use the bsearch_spell_dictionary_batch_size filter to change the batch size.

add_filter( 'bsearch_spell_dictionary_batch_size', fn() => 500 );

A rebuild fills a shadow table and swaps it in when it is complete, so the existing dictionary keeps serving suggestions for the whole of the rebuild. Earlier versions emptied the table first, which left “Did you mean” returning nothing until the rebuild finished.

Words that differ only by letter case or accents are combined into a single entry with their frequencies added together, matching the dictionary table’s collation. Rebuilding no longer fails when your content contains such pairs.

Developer Filters

  • bsearch_spelling_suggestionOpens in a new window — Filters the final suggestion resolved for a single search token, after the search log, content index, and enchant fallback have all been tried.
  • bsearch_did_you_mean_min_searchesOpens in a new window — Filters the minimum search-log count required for a term to qualify as a suggestion, overriding the Minimum searches to qualify as a suggestion setting.
  • bsearch_did_you_mean_enchant_locale — Filters the locale (default en_US) used to request a dictionary from enchant for the fallback.
  • get_bsearch_did_you_mean / get_bsearch_autocorrect_notice — Filter the generated markup for each notice.
  • bsearch_spell_dictionary_batch_size — Filters the number of posts read in each dictionary rebuild batch. Default 2000.

Important Considerations

  • Zero-result searches only: Suggestions are only computed when a search returns no results — there’s no performance impact on normal searches.
  • Auto-correct is re-run, not guessed: In Auto-correct mode, the corrected phrase is only adopted if re-running the search with it actually returns results; otherwise Better Search falls back to showing the normal zero-results page.
  • Accepted corrections reinforce your search log: When Auto-correct mode succeeds, the corrected term is logged as a search, strengthening future suggestions.
  • Repeated saves no longer skew frequencies: When you save a post, its title words are added to the dictionary only if they are missing, rather than having their counts incremented again. Frequencies now reflect how common a word is, not how often a post was edited. The scheduled rebuild recomputes true frequencies twice a day.

Available to AI assistants

When this feature is enabled, Better Search Pro also registers the better-search/suggest-search-term ability, so a connected AI assistant can ask your site for a correction after a search returns nothing. Switching the feature off removes the ability entirely, rather than leaving a tool that always answers with nothing. See Better Search Abilities API.

See also

Was this article helpful?

Tagged:

Related Articles