A few years ago I started playing with what was then a revolutionary new tool. It was called Podzinger, and it combined unique speech recognition software (one that is supposedly also used by the Central Intelligence Agency in the U.S.) with an indexing service to offer a search engine for podcasts. Podzinger was able to "listen" to an audio file and then take me to the section mentioning my search term.

So if I'd have searched for podcasts mentioning Steve Irwin's fateful last meeting with a stingray back in 2006, Podzinger would have returned a few dozen results of podcasts mentioning the accident - and the results would have specified exactly where, within the audio file, the words "Steve Irwin" and "stingray" were said.

Since its launch, Podzinger has expanded, adding staff and even landing a new name (EveryZing). You can still use EveryZing to search for content, but it now comes as a tool that can live on a Web site to offer multimedia search.

I recently spoke with Tom Wilde, EveryZing's CEO, about the future of audio/video search and the mistakes many publishers make when putting their video content online.

Amy Webb: What common mistakes do you see journalists make when it comes to audio/video content online?

Tom Wilde: (paraphrased) A lot of a/v content is not well-tagged. Scale is also an issue, because not everyone thinks about the video being archived and searched on site. At some point, you can't just offer a long list of all the videos you've created - users expect to be able to search for a video and get an accurate result.

Content needs to be authored so that it's discoverable on the Web. Publishing systems aren't set up using SEO best practices, and because of that, many news organizations continue to struggle with SEO. And their consumers have trouble with usability.

AW: Is that because users have a certain level of expectation once they arrive at a Web site?

TW: Absolutely.

AW: EveryZing indexes audio and video files, and I know that some news organizations, such as Fox Sports and Boston.com are using EveryZing as the multimedia search engines on their sites. Do you think that there's a disconnect between how journalists think people are searching the Web and how most people actually are searching for video and audio files?

TW: (paraphrased) We're in a single search box environment, where users expect to be able to type in one set of keywords and receive everything they need. That's Google. You type in a set of keywords and get everything you need. Audio and video that's offered on a news site should be easy to find. It doesn't help the user to simply list all of your multimedia content in one place. It should be indexed and ranked so that a user receives the same detailed, accurate search results as they would looking for a text story.

AW: What does that mean for the kind of video that's being offered on news sites? With services such as EveryZing, Truveo and Mefeedia launching, do journalists need to think about how they're putting together their videos?

TW: As journalists shoot videos and podcasts, they must be mindful of how those files will eventually be found. If people speak in shorthand or jargon, that causes problems for the services that index that content. Videos with professional production quality and clean audio are the easiest for EveryZing to find.

(paraphrased) Going forward, journalists should try to put together short Web videos with search engines in mind. Videos should offer very clear speech and should include good tags and other metadata.

Watch a Boston.com video of Tom Wilde talking more about how EveryZing works. Tom talks more about SEO in this video.