Add-on: AI Search
$79/year
Your back catalog is worth more than keyword search can deliver.
Search your entire back catalog in everyday language.
Let your audience — or you — find exactly what they’re looking for, using language they understand.
A subscriber who joined last month has no idea that episode 84 covers exactly what they’re looking for right now — and keyword search won’t help them find it unless they already know the right words. “What do I do when I’m burned out?” doesn’t match an episode called “Managing Creative Fatigue.” The content is there. The subscriber wants it. Standard search can’t bridge the gap.
AI Search works in everyday language. Semantic understanding means the system finds the right content even when the words don’t match — it understands what the question means, not just what it says.
You bring your own API key.
You connect your own account directly to an AI provider of your choice. Benecaster passes search requests through to that account and returns results. You pay the provider directly at their standard rates. Benecaster doesn’t add a margin, a service fee, or a per-query charge. The cost of running AI Search on your site is whatever the provider charges — nothing more.
Supported providers
- OpenAI — text-embedding-3-small, text-embedding-3-large
- Google Gemini — gemini-embedding-001
- Cohere — embed-english-v3.0, embed-multilingual-v3.0
- Anthropic — Claude Haiku 5.5, Claude Sonnet 5.5, Claude Opus 5.5
You choose the provider and model once, on Settings → AI Search, and they serve every show.
Running a model of your own, on your server or your network? A developer can add it to the list as a provider of its own. See Add your own search provider to AI Search.
Not only English
AI Search isn’t limited to English. Google Gemini’s embedding model covers more than 100 languages, OpenAI’s embedding models are multilingual, and Cohere offers a multilingual model alongside its English one for a show in another language. Your listeners search in the language your show is in.
How it works
AI Search works one of two ways, depending on the provider you choose.
With OpenAI, Google Gemini or Cohere, AI Search converts your episode content — titles, descriptions, and show notes — into mathematical representations called vector embeddings, once, in the background. These capture the semantic meaning of the text, not just the words it uses. When a subscriber types a query, their question is converted to the same kind of embedding and compared against the index. The results are episodes whose meaning is closest to what they asked — regardless of whether the actual words match. Indexing costs a little once per episode, unchanged episodes are never sent again, and each search after that costs very little.
With Anthropic, nothing is indexed ahead of time. Each search sends the question and your show’s episode list (each episode’s title and a short summary) to Claude, which reads them and ranks the episodes that answer it. There’s nothing to build first, but each search costs more than with an indexing provider.
That distinction matters at scale. A subscriber searching for “interviews with founders” finds those conversations even if every episode is titled “A conversation with…” A search for “episodes about systems thinking” surfaces content that covers the concept whether or not that exact phrase appears anywhere. Your entire back catalog becomes searchable through natural human language — the way subscribers actually think about what they want, not the way episode titles happen to be worded.
Search covers episode titles, descriptions, and show notes. Add Transcription Service and, with OpenAI, Google Gemini or Cohere, it searches what was said in each episode too, for every listener allowed to see that episode’s transcript.
For your audience. And for you.
Backend search panel for podcasters
AI Search adds a search panel beside every Benecaster admin screen — open it from the Dashboard or with Alt+Shift+S while you’re recording or prepping an episode. Type a topic or question in your own words and get episode references without leaving the backend: episode number, title, publish date, and a matched excerpt. Searches episode titles, descriptions, show notes, and the transcripts Transcription Service holds, whether transcribed or imported.
“Did I already cover X?” is a question every podcaster asks, and digging through a back-catalog manually mid-recording is painful. Type the concept, get the episode — in seconds, from the admin screen you’re already on. When Guest Manager is active, the panel also finds your guests: by name, organization or title, or as the guests on the episodes that match your topic, each with their latest episodes.
Subscriber-facing widget
Place AI Search anywhere on your site — sidebar, full-page, or modal. You control how results appear for visitors who aren’t subscribed or don’t have access to a particular tier.
Result visibility is configurable per widget instance:
- Hidden (default) — inaccessible episodes don’t appear in results at all. Clean results, no friction.
- Locked — inaccessible episodes appear in results with a locked card. The title is visible; the description and any transcript excerpt are withheld. A lock icon, tier label, and upgrade CTA appear in place of the content.
The locked mode turns search into a conversion surface. A visitor searches for something relevant, sees that you have an episode about exactly that, and hits a “Subscribe to access” prompt right there in the result. The CTA destination and label are both configurable. The benecaster_ai_search_locked_result filter lets developers replace the locked card markup entirely.
Because visibility is set per widget instance, the same site can run locked mode on a public search bar and hidden mode inside a subscriber-only premium search area — independently, with no conflict.
You can also gate the search interface itself by tier. A common pattern: Standard Search on the public episode archive for all visitors, AI Search as a premium feature for paid subscribers only. Both shortcodes — [benecaster_search] and [benecaster_ai_search] — can coexist on the same site, pointing at different contexts with no conflict.
Or keep AI Search for signed-in visitors only. Everyone else gets Standard Search in its place, free to run, so bots never reach your provider bill. It’s off by default, and one setting covers the whole site.
Episode correlation and navigation
AI Search identifies semantic relationships between episodes across your back-catalog and adds a new way to choose related episodes: by meaning, not just by shared tag, season or guest. Pick it once for the Related Episodes list on your episode pages (Which episodes to list on the show’s Templates tab), or use it anywhere with [benecaster_related_episodes query="by_semantic"], and every episode page suggests the episodes closest in subject. With Guest Manager, you can also have it prefer episodes that share a guest. Episodes a visitor can’t access show with a lock, the same as in any related list. Keeps listeners in your library rather than back in their podcast app, improves retention, and creates internal linking signals that benefit SEO.
Your plan already includes Standard Search. Here’s when AI Search goes further.
Standard Search is built into every Benecaster plan at no cost — full-text keyword matching across episode titles, descriptions, and show notes. It covers the majority of searches: episode number lookups, guest name searches, exact topic matches.
AI Search is worth the additional cost when your archive is large enough that keyword misses become meaningful, or when you want subscribers to browse by concept rather than title.
| Standard Search | AI Search | |
|---|---|---|
| Included in plan | ✓ Free | $79/year |
| Search type | Keyword matching | Semantic / natural language |
| API key required | No | Yes |
| “Productivity while traveling” | ❌ unless those exact words appear | ✓ |
| Exact episode title or number | ✓ | ✓ |
Even More Powerful With
- Transcription Service — With OpenAI, Google Gemini or Cohere, transcript text is included in the search index alongside titles and show notes, and a listener’s search matches it whenever they’re allowed to see the transcript. For long-form interview shows, this is the difference between AI Search finding the right episode and missing it entirely — most of what was actually said lives in the audio, not the description.
- Guest Manager — When Guest Manager is active, the admin backend search panel also finds your guests — by name, organization, title or bio, or by topic through the episodes they appear on, even guests you’ve booked who aren’t on an episode yet — and shows which episodes each one appeared in, all from the same interface. Related episodes can also prefer episodes that share a guest.
- Guest Manager and Transcription Service together — Label who’s speaking in your transcripts, match each speaker to a guest’s profile, and the search box on your site can answer “what has [guest] said about [topic]?” across your entire archive, from the guest’s own words rather than an episode description. Passages appear only when the guest actually spoke about the topic, and listeners only see passages from transcripts they’re allowed to read. Needs all three add-ons on for the show, and OpenAI, Google Gemini or Cohere as your provider.
- Blocks & Widgets Library — Place the search box on any page as a block or an Elementor widget, with no shortcode to type: as a sidebar, a full page or a modal.
- Podcast Workflow — A search box in the Recording View finds your past episodes mid-conversation, for the “as we talked about back in episode 40” moment.
The three add-ons work best together: Guest Manager gives guests an identity, Transcription gives their words attribution, and AI Search makes it all findable — by topic, by concept, or by speaker.
Requires
- Benecaster with any plan
- An API key from a supported provider (OpenAI, Google Gemini, Cohere, or Anthropic)
- To search what was said in your episodes, not just the descriptions: Transcription Service (transcribed or imported transcripts) and OpenAI, Google Gemini or Cohere as your provider. Listeners’ searches match a transcript only when they’re allowed to see it