Should Podcasting Be Part of Your Learning Mix?

About this paper

I make learning podcasts. I have been doing this since 2019, and I now have four of them, adding up to more than two hundred episodes. I also advise organisations on whether to make their own, as part of their learning mix, and a fair proportion of the time my advice is that they should not.

This paper sets out what the research actually supports about learning through audio, what it does not support, and what has changed in recent years. It ends with ten key questions you can answer to assess whether audio-first learning is a fit for you. There is no form to fill in, no follow-up sequence. I’ve created this paper to be read, argued with, and if you think it might be of use to others, passed on.

Contents

1. The most important question people don’t ask Framing the decision as a trade-off inside an existing learning mix, rather than a question about podcasting on its own.

2. What the medium does that the others don’t Four properties – reach, trust, intimacy and emotion – with current listening figures for the UK and US, and what the shift towards video does to the argument.

3. What the cognitive science supports, and what it doesn’t Working memory, the modality principle and cognitive load; the transient information effect and what it predicts audio will handle badly; and two claims about audio that the evidence does not support.

4. Why dialogue works The case for the interview format in particular, from Socratic questioning to current accounts of thought as improvisatory.

5. Where it fits in your learning mix Podcasting mapped against the common training categories in L&D, and against the frameworks used in higher education and in vocational education.

6. What it costs What has become cheaper since 2023, what has not, and whether to generate a show with AI.

7. If you’re doing it already and it isn’t working Seven common failure modes, and what usually lies behind them.

8. Ten questions A diagnostic to work through, with a note under each on what a weak answer indicates.

9. If the answer is yes Where to take it from there.

References

Sections 2 and 3 make the evidential case. Section 5 is the one most readers will want if they have already decided in principle and are working out where a podcast would sit.

1. The most important question people don’t ask

Most conversations about podcasts skirt around an important issue. I think it’s the most important issue. Someone asks whether the organisation should make a podcast, and the discussion moves straight to microphones, hosts and episode length. Too often, the key question doesn’t get asked: does audio do something your existing provision doesn’t, and can it do that thing well enough to justify the budget you’ll need to pay for it?

After all, most organisations nowadays offer multimodal learning – and some provide a dizzying array of content types and training formats; elearning, live sessions, video, coaching, a resource library, a leadership programme … and whatever else has accumulated. Each of those does a job, some of them well, some badly. So if you’re going to add a new modality to the mix, it has to earn its place alongside this battery of learning methods – and in almost every organisation I have worked with, that means doing a bit less of something else.

So here’s how you might frame your key decision: not “should we make a podcast” but “what would a podcast displace, and would that trade-off increase the value we offer to our learners?” That’s the question to answer right at the start.

Here’s my view, for what it’s worth. Audio does something the rest of your mix can’t do, and definitely can add value to your learners if the situation is right. However, the case for it is often overstated by people who sell podcast services, and it really pays to be clear about what it can and cannot do.

The following, based on 25 years in learning, seven as a professional podcaster, and eight podcast seasons digging into learning theory with my co-host and mentor Donald Clark, is a warts-and-all survey of the facts. I’ll lay out the evidence that supports the effectiveness of audio-first learning, as well as what that evidence does not support. Where two good sources disagree, I try to work out why.

The last section poses questions which, by then, you should be able to answer without my help.

2. What the medium does that the others don’t

When I first came across podcasts as a listener, I was blown away by how knowledge-rich and ‘sticky’ they seemed; how informative and habit-creating. But here’s the observation that first made me curious about their potential power as a learning medium.

I noticed that professionals who rolled their eyes at a twenty-minute elearning module (and usually bailed half-way through) showed a very different attitude to podcasts. Not only would they stay the course for a forty-minute podcast, they would then look for the next one. They would binge whole series, or follow a weekly release as a matter of routine. At a time when people were telling me all the time that human attention spans were decreasing, I saw quite the opposite. A 15-minute module of ‘micro-learning’ on time management? Nein danke. A 4-hour podcast on Middle-East politics? Count me in.

That is not a failure of discipline on the professionals’ part. The two formats are doing different things. But harnessing the power of podcasts to capture and hold attention for difficult subject matter seemed to me like something that could be a massive win for learning.

Here are the four reasons that – to my mind – explain what the power and appeal of podcasts is all about, and why we should work hard for their inclusion in the learning mix.

Reach

Podcast listening is now a mainstream habit. Also, it is no longer a young person’s habit alone, but something for virtually anyone of working age.

In the UK, 27% of adults listen to a podcast every week, according to RAJAR figures reported in Ofcom’s Media Nations 2026. A decade ago that was 7%. The 45–54 age group now has higher-than-average reach, and 13% of people over 65 listen weekly.

In the US the numbers are larger. Edison Research’s Infinite Dial 2026 puts monthly podcast consumption at 58% of the population aged 12 and over, and weekly consumption at 45%. The sharpest movement is at the older end: monthly online audio listening among Americans over 55 went from 52% in 2024 to 70% in 2026.

Let’s keep this in proportion, however. Ofcom’s same report has live radio reaching 86% of UK adults weekly and accounting for 66% of all time spent listening to audio. Podcasting is a significant medium: it is not the dominant one – and anyone who tells you otherwise is probably trying to sell you something.

What matters for learning is where the listening happens. Audio occupies attention that is otherwise unused – commuting, driving, walking, exercising, doing something with your hands. No screen-based format can reach into that time. If you have a workforce that is distributed, mobile or not desk-based, bearing in mind that some 80% of the workforce is probably not desk-based (a widely circulated figure originating with Emergence Capital, though the methodology is not public) this makes a strong argument for podcasts as part of your learning mix.

Trust

Listeners report unusual levels of trust in podcast hosts. Acast’s Podcast Pulse 2025 found that when listeners are considering a purchase, podcast hosts rank level with journalists for credibility at 33%, ahead of YouTubers at 31%, social media influencers at 28% and celebrities at 25%. More striking for our purposes, 84% of listeners said a podcaster had changed their mind about something they once believed.

Two qualifications, both of which matter in an age where trust, in general, is on the wane.

The survey base is podcast listeners, not the general public. It tells you that people who listen to podcasts trust podcasters. It does not tell you that the population at large trusts podcasts more than it trusts the BBC. Every trust statistic in circulation on this subject has the same limitation, including the ones I have used previously in conference talks.

We should also note that Acast sells podcast advertising. The study was designed to move ad budget. The research was conducted by an independent agency across nine markets with more than 2,500 respondents, so it is not worthless, but it is not disinterested either.

There is a further qualification the research cannot help with. Trust in the medium does not automatically transfer to your organisation’s podcast. An internal show made by the communications team inherits whatever credibility the communications team already has. If your people are sceptical of official channels, putting the official channel into their ears does not fix it, and may make it worse. If, on the other hand, you are truly confident that you have a good level of trust within your workforce or student body – then crack on. And if official channels lack credibility, then what about unoffical ones – something closer to the grass roots of the organisation or institution?

Intimacy

Cultural anthropologist Edward T. Hall coined the term ‘Proxemics’ in 1963 as the study of how people organize and feel about personal space. His later work defined four distinct ‘distance zones’, from ‘Intimate’ (0-18 inches) to ‘Public’ (12 to 25+ feet). Hall’s zones are a descriptive framework from 1966 rather than a measured finding, but they are broadly consistent with later work on interpersonal distance preferences. Podcasts inhabit the intimate space Hall speaks about, which most people reserve for close loved ones (lovers, children) and wrestlers for their opponents. A voice speaking at such close proximity for forty minutes builds a relationship of a very different kind from that which you might form with a speaker in a lecture hall, or a trainer in a training room.

This effect has a name and a research literature going back to Horton and Wohl’s 1956 work on parasocial interaction, describing the one-sided sense of familiarity audiences develop with broadcast performers. Podcasting is close to the purest form of this effect, because there is very little else to get in the way – no set, no graphics, no cutaways. Just someone talking to you.

Neil MacGregor, former director of the British Museum, became a familiar voice to the UK public through his popular BBC radio series (now an extremely successful podcast), A History of the World in 100 Objects. Talking to Roman Mars on the 99% Invisible podcast in July 2026, he described how strange and disconcerting it was to be recognised by a stranger solely from his voice: ‘The first time it happens, it’s very, very peculiar.’

Though much less famous, I have had the experience at conferences of being greeted like an old friend by complete strangers who feel that they have come to know me pretty well through listening to my podcasts. It’s extremely gratifying, of course. But yes, also peculiar.

For learning, the useful consequence is that people will accept from a familiar voice things they would reject from a corporate deck. This throws a lot of importance on the choice of host. In the organisational context this could mean using someone who is well known in the company, or the industry or practice area – or perhaps an external celeb. Alternatively, it could just mean building familiarity through regularity and consistency. And not using hosts the timbre of whose voices have the same effect as dragging fingernails down a blackboard.

Emotion

Audio carries emotion with unusual efficiency, and emotion has a real bearing on attention and on what people remember. Studies from the 1990s (notably, Cahill & McGaugh, Nature 1994 371:702–704) have shown that emotional arousal aids retention over time, with the advantage appearing at a delay rather than immediately. In a sense it’s immaterial whether the emotion aroused is pity, fear, intrigue, excitement, amusement … or whatever: the audio just has to have an emotional effect. Activation is what matters. We could state this baldly in three words: don’t be boring.

However, it’s important that the emotion aroused, whatever it is, relates strongly to the point of the story. There is a danger otherwise that learners remember details, but forget the message. Emotional arousal narrows attention, sharpening memory for what is central and blurring out what isn’t (Dolcos & Denkova, 2014). The lesson isn’t ‘make it emotional,’ it’s ‘make it matter’.

The practitioner-theorist Nick Shackleton-Jones, often mentioned in my podcasts (and twice interviewed as a guest) speaks and writes eloquently of how these psychological effects should influence learning design. Nobody can learn anything about a subject if they don’t care about it, he says. If a subject relates closely to things the subject cares deeply about already (e.g. earning more money, paying less tax) they will expend considerable time and effort extracting everything you know about it. If on the other hand it’s a subject they care less about, such as regulatory compliance, you might have to work a lot harder to unpack the emotion around your subject and link it to something they do care about.

Podcasting is primarily a story-telling medium, and stories are the most common means we all use to attach emotion to the information we have to convey. Doing that goes way back. It’s a primal, even biological inheritance. Oracy predates literacy by a matter of millennia. Poetry, with its mnemonic scaffolding of metre, rhyme, epithets and the like, is much much older than prose. And voice is among the first things that give us information about the world we have joined when we are born. The voice of a mother is beginning to teach first lessons about life before infant eyes are even able to focus properly.

Small wonder then, that so much of our emotional landscape can be terraformed by individual voices. You can fall in love with your ears. A voice can inspire you to a greater resolve (Churchill or FDR in their wartime radio broadcasts for example), or contempt (Lord Haw Haw, same). Certain voices inspire laughter (Mel Brooks, or for UK listeners, Victoria Wood), others wonder (David Attenborough, Carl Sagan). Some voices have an ASMR quality (almost anyone from the south of Ireland), or a reliable ability to spark instant irritation (pick your least favourite politician). And nothing brings us sadness like the voice of a deceased loved one recorded on an answerphone, or holiday video.

Video killed the radio star?

Having said all this, there is a fact about the podcast medium that any paper arguing the case for audio in learning should at least acknowledge.

Video podcast consumption overtook audio-only in the third quarter of 2025. Edison’s Podcast Consumer 2026 reports that 82% of weekly podcast consumers actively watch video, against 78% who listen to audio only. As recently as 2023 the split was 89% audio-only against 73% video.

Acast’s data points the other way. It finds that 40% of listeners mostly listen and sometimes watch, 27% do both, and only 17% consume video exclusively.

Both can be true, because they measure different things. Edison counts anyone who watches at all against anyone who listens audio-only. Acast asks listeners which mode they mostly use. Acast also monetises audio feeds, which gives it an interest in the answer.

However, what this means for you is that “podcast” no longer reliably means “audio”. The medium is now dual-format, and the reach argument above applies to the listening portion of it, not to all of it.

It does not undo the case. The properties that make podcasting useful for learning belong to the form rather than to the delivery channel – long, conversational, unhurried, chaptered, and comprehensible without the visual element. A show that is built audio-first will work in a car and also on a screen. A show built as a video first, with slides and cutaways that carry meaning, does not work in the car, and you have, basically, made a different product. It’s not a podcast.

The podcast, even on YouTube, is an audio-first medium.

3. What the cognitive science supports, and what it doesn’t

This section attempts to bring a bit of rigour to what the science really tells us about audio-first learning. Much of what circulates about audio and learning is a thin layer of real research under a thick layer of things people wish were true, and some of the citations in general circulation do not say what they are claimed to say. The truth is that there’s some nuance of which we have to take account. Here is what the theory does support, and what it doesn’t.

What it supports

Working memory has separable channels. Baddeley and Hitch’s model, first set out in 1974, describes a phonological loop for verbal and acoustic material and a visuospatial sketchpad for visual and spatial material, coordinated by a central executive. Audio uses one of these without competing for the other. This is why listening while walking is comfortable and reading while walking is not, just apart from the danger that with your nose in a book you will likely bump into a tree.

The auditory channel carries explanation well. Richard Mayer’s modality principle holds that people learn better from narration combined with graphics than from on-screen text combined with graphics, because printed words and pictures compete for the same visual channel while spoken words do not.

Audio-only avoids split attention. John Sweller’s cognitive load theory distinguishes load intrinsic to the material from load created by how it is presented. The split-attention effect – forcing a learner to integrate two sources of information presented separately – is one of the commonest faults in screen-based learning design. Audio-only material, by definition, can’t commit this error. And audio-first material is much less likely to.

However, here’s where the findings get a little nuanced. In experiments reported in Applied Cognitive Psychology, Leahy and Sweller found the modality effect reversed when the spoken text was long: written text plus diagram beat audio plus diagram, contrary to most previous modality research. Their explanation is that lengthy transitory auditory material overloads working memory in a way permanent written text does not. In the same journal, Singh, Marcus & Ayres (2012) found that segmenting spoken text reduces the transient information effect. There are certainly aesthetic reasons for dividing a podcast into ‘segments’ or ‘chapters’, broken up by music and sound design, but the research literature would seem to give strong backing to such an approach: it is an evidence-based recommendation rather than a preference.

Anecdote time: the redundancy effect was one of the first things I ever learned about content creation. I wasn’t aware of it as a scientific principle at the time. I was in an edit suite for a corporate video I had scripted and there was one sequence that was just not cutting the mustard. The producer explained why to me (rather tetchily, I thought). I had specified some copy to run over a talking head who was effectively making the same point, but not in anything like the same words. ‘Text and audio will fight,’ he said. He was not a cognitive psychologist, nothing like: he produced videos – but many years of experience told him it wouldn’t work.

What it does not support

Mayer is not an argument for audio. His multimedia principle is that people learn better from words and pictures together than from words alone. Read straight, that counts against audio-only delivery for anything with a spatial or visual component. The modality principle is a finding about how to combine channels in multimedia, not a finding that audio beats video. It is easy to read it as the second thing. It is not.

Transient information is the real limit. Spoken material disappears as it is delivered. The listener cannot slow it down, re-read a sentence, or scan back to the definition three paragraphs earlier without effort that itself costs attention. Sweller and colleagues describe this as the transient information effect, and it predicts what audio is bad at with some precision. Podcasts mitigate this with the ‘back’ button that let you repeat the previous 15 or 30 seconds. Personally I use that button a lot – but a learning designer can’t necessarily rely on the learner to make active use of it. Particularly if they are doing the ironing at the time. There is also research showing that where people listen measurably affects how engaged they are (Harrison et al., 2023).

Audio handles ideas, arguments, judgement, worked reasoning and stories. It handles procedures, sequences, numbers, definitions and anything that must be got exactly right less well. If your content is a compliance procedure or a product specification, you have to take that on board. Where a compliance requirement tends to encounter grey areas, or is regularly being updated due to changing circumstances, a regular podcast feed is a great way of exploring nuance and keeping knowledge up to date. But where procedural information is centre-stage, it will tend to work best as additive and complementary to existing media, rather than as a replacement.

Audio does not produce better long-term retention. This claim is widespread and I have repeated versions of it myself (mea culpa). In my case I attributed it to a 1977 paper by Baddeley and Hitch. That paper is about the recency effect in free recall and makes no such modality comparison. I can find no good evidence for a general auditory advantage in long-term retention.

The education literature is mixed. Heilesen (2010) reviewed 2004–2009 and concluded that measured against student performance the evidence is fairly weak, with a general absence of longitudinal studies, while podcasting nonetheless improves the study environment and students use it rationally as a supplement. Studies also find, consistently, that effectiveness depends on the quality of the content, the motivation of the learner and how well the material is integrated into everything else. You can’t just do it, you have to do it well. In fact, this statement might as well stand as a summary for the whole subject. The medium does not do the work. What you put in is what makes it instructionally effective.

A note about length

Carvalho et al. (2009), cited in the Thomas review, recommend that podcasts conveying detail and facts should not exceed 30 minutes, on the grounds that long podcasts cause loss of attention and reduced comprehension. Earlier in this paper we talked about forty-minute conversation, and that is the length I generally aim for in my own podcasts. However, if we recall the point about chapterization, and accept that a podcast episode might contain front matter, housekeeping, recapitulation, outro credits, etc. (and in a commercial podcast even ads) it’s quite probable that the detail and fact portion of the podcast may only be 30 minutes in total. However we should also bear in mind the distinction between conveying detail and conveying ideas made elsewhere in this paper.

4. Why dialogue works

Almost everything above concerns audio as a delivery channel. But we shouldn’t overlook the role of dialogue, which is one of the most potent aspects of the podcast as a learning medium. Many fine learning podcasts take the form of monologue – but far more are couched as conversations. And here, this thoroughly modern form has ancient forebears.

Questioning surfaces reasoning that exposition hides. This is the oldest observation in the pedagogical literature – it is what the Socratic dialogues are for – and it survives translation into the podcast medium. An expert writing an article produces conclusions. The same expert being interviewed produces the reasoning behind them, including the parts they are unsure about, because a good interviewer keeps asking why.

The interviewer’s real function is to stand in for the audience’s ignorance. I can ask the question your people are too junior, too polite or too busy to ask, and I can refuse the first answer if it arrives in jargon. That is a service the format performs which a written brief can’t.

There is a deeper version of this argument. Nick Chater’s account of cognition in The Mind Is Flat describes thought as improvisatory – the mind generating and completing stories, rather than reading off settled beliefs. If thinking is closer to talking than to reading, then hearing someone think aloud is not a degraded substitute for reading their conclusions. It is a more direct view of the thing you are trying to teach.

Reading evolutionary psychology can give you an interesting take on Mayer’s modality principle. The human brain is really bad at remembering lists, but really good at remembering paths through a landscape. Writing was invented at least in part to repair this deficit: some of the earliest forms of writing were for administrative and record-keeping purposes. Lists. A list is a type of spatial information, and lists are best presented visually (e.g. bullet points on a Powerpoint).

But if we look even further back, to the beginnings of language, we might imagine a bunch of hunter-gatherers in search of avocados for their breakfast – millenials from a far off millenium – dispatching a member of the collective to go and find out where they are growing, and that individual, having found some, coming back to tell the rest of them how she got to where they are - right at the big tree, left by the waterhole, and so on. That is something like what we’re doing with podcasts, and it is a very primal type of learning. The podcaster is basically saying: I’ve explored this landscape of (insert knowledge area here) and now I’m going to tell you how the land lies. It’s important to say that this is my interpretation, not an experimental finding!

… And one caution on the Socratic point. Recorded dialogue is not, strictly speaking, Socratic. Socrates questioned the person in front of him and adjusted his questioning based on the answers. A podcast cannot respond to the listener, so what it offers is the experience of hearing someone else being questioned. However, truth to tell, that’s no less a learning experience than reading Plato’s Republic, which is itself a written record of Socrates questioning people (or Plato’s reconstruction of it, at any rate) – not that I would ever compare my own interviewing skills with those of the great Athenian!

5. Where it fits in your learning mix

There are significant differences, for our purposes, in the organisational and educational contexts, so I’ve treated each area separately below.

One important caveat first, however. The mappings from learning theory to podcast suitability given in this section are my reasoned judgement applied to established frameworks, rather than empirical findings. I’d expect informed people to disagree with some of them. But I wanted to provide some guidance, rough-and-ready as it might be, to help people focus in on where podcasting can add most value in the mix.

L&D

Research from 2026 carried out by TalentLMS, a vendor, yields this list of the most common types of training currently being given in organisations:

• Compliance – the most widely delivered training, reaching 66% of employees in 2025
• Leadership and manager development – the number one HR priority for the third consecutive year, 64% making it a 2026 priority
• Digital and technical skills – 73% of HR managers rank this their main focus for 2026
• Upskilling – 57% of employees, up from 50% in 2022
• Soft skills – 47% in 2025, up from 38% in 2024
• Product training – 45%
• Onboarding, health and safety, DEIB, change management

So which of these pressing needs is podcasting most able to fulfil? Spoiler alert: not all of them. A quick dip into the learning theory most familiar to L&D gives us some pointers.

Gagné (1965) – made the point sixty years ago that different types of learning require different instructional conditions. He recognised five categories of learning outcome: - Verbal information - Intellectual skills - Cognitive strategies - Attitudes - Motor skills Podcasting is strongest on attitudes and cognitive strategies. Interestingly, these are the two that formal e-learning handles worst. With the others in this list audio-first learning is of partial, or split use, and motor skills are a no-no.

Anderson & Krathwohl (2001) - revised Bloom’s 1956 taxonomy, adding a knowledge dimension of four types: - Factual - Conceptual - Procedural - Metacognitive. Here, podcasting is strongest in conceptual and metacognitive, and weak in factual and procedural.

Mapping these considerations to training types gives us the following table.

We could also add to this list some more specific use cases:

Before and after a programme. Building appetite for something before it starts, and keeping the ideas alive once the cohort has dispersed.

Shared reference for distributed people. A common set of ideas and vocabulary across sites, territories and time zones.

Reaching the non-desk workforce. Field engineers, drivers, care staff, retail, anyone whose learning currently depends on being pulled off the floor.

Niches too small to justify a course. A subject with fifty interested people in the organisation is not a business case for an e-learning build. It could well be a good business case for six podcast episodes.

Communities of practice – A podcast and its listeners can function as a community of practice in Lave and Wenger’s sense – people with a shared concern who deepen their understanding through ongoing interaction. Note that a community of practice requires interaction, however. If you want the community, you have to build the thing around the show that lets people talk back – a channel, a forum, a regular session.

It is worth noting, as regards our next section, that Lave & Wenger’s theory of legitimate peripheral participation was derived from studies of apprenticeship. It is therefore just as relevant, if not more so, to communities of practice in VET.

Higher and vocational education

Corporate L&D, higher education and VET use different vocabularies to discuss these questions. Here’s a version of the above for readers on the education side.

Higher education

HE does not think in training types. It thinks in constructive alignment – Biggs and Tang’s framework, dominant in UK, Irish, Australian and much European quality assurance. Intended learning outcomes drive teaching and learning activities, which are matched by assessment tasks.

The practical consequence for a podcast is worth saying bluntly. In an aligned design, anything not connected to an assessed outcome becomes optional, and optional material is what students drop first. If you want a podcast used in HE, it has to be tied to something that is assessed, or explicitly positioned as enrichment and resourced accordingly. This is the single most common reason HE podcast projects fail, and it is a design failure rather than a content one.

S.O.L.O.: Biggs & Collis (1982), the structure of observed learning outcomes: - Prestructural - Unistructural - Multistructural - Relational - Extended abstract. Used in HE as a more defensible alternative to Bloom because it describes the structure of a response rather than assuming a hierarchy of processes.

Podcasts do the top two well. Relational understanding – connecting ideas into a coherent whole – and extended abstract – transferring beyond the case in hand – are what a good conversation between experts produces. The lower levels, accumulating discrete facts, are the format’s weak ground.

Threshold concepts: Meyer & Land (2003 onwards). A threshold concept is one that transforms how a learner sees a subject once crossed. Their characteristics: troublesome, transformative, irreversible, integrative, bounded. The state of being stuck part-way is liminality, where a student oscillates between an earlier, cruder understanding and the one their tutor wants.

Three reasons this suits audio-first: 1. A threshold concept is by definition one where written explanation has already failed. If the textbook worked, it would not be troublesome. 2. Hearing an expert work through the difficulty – including saying that it was hard for them, and where they went wrong – models the process rather than the outcome. That is close to what tutorials do and expensive to scale any other way. 3. Liminality is uncomfortable and isolating. Students in it generally believe they are the only one. Hearing other people describe being stuck normalises it, which is an affective intervention on a cognitive problem.

Studies of podcasting in higher education consistently find that supplementary use outperforms substitutive use (Thomas et al. 2025, 91 publications). Podcasts alongside teaching help. Podcasts replacing teaching produce mixed to poor results. The established use case is pre-class input in a flipped design.

McGarr, O. (2009) supports the distinction between using podcasts in a substitutional, supplemental or creative role. McGarr’s point is about how each positions the learner: substitutional makes students passive recipients; supplemental can either reinforce passivity through revision material or produce deeper engagement through exposure to varied perspectives; creative makes them knowledge producers.

Vocational education

A note on terms before we start. I use VET – vocational education and training – throughout, because it is the standard term internationally and the one my European readers will expect. In the UK, most of this provision sits in FE colleges, though FE is a sector rather than a type of learning, and it also covers A-levels, access courses, ESOL and adult and community education. The two are not interchangeable.

VET in Northern Europe is a larger, better-funded and considerably higher-status system than its UK equivalent, as readers in Denmark, Germany, the Netherlands, Austria and Norway will be aware. The systems are dual or alternating by design: German, Austrian and Swiss duale Ausbildung splitting apprentices between Berufsschule and employer; Danish vekseluddannelse in the EUD programmes; Dutch MBO with its school-based BOL and work-based BBL routes; the Norwegian 2+2 model.

EQF – European learning professionals also have an up-to-date, broadly accepted taxonomy to work with here: the European Qualifications Framework, supported by Cedefop, the EU agency for VET.

The EQF defines eight levels entirely in terms of learning outcomes:

The ‘responsibility and autonomy’ category comprises judgement, professional conduct, knowing what to do when nobody is telling you. It is the hardest thing to timetable, and it is learned largely by exposure to people who already have it. A podcast interview series is exposure to people who already have it.

One problem VET research is preoccupied with is the school–workplace gap. Connecting what is taught at a vocational school with what happens at the employer is a central and persistent problem of the field. Guile and Griffiths’ connective model of work experience is the standard reference point.

A podcast is a plausible way of supporting students across this boundary. The same material can reach both settings, and it fits into time that exists in both – travel, workshop downtime, the commute – without either institution having to give up timetabled hours.

Final thoughts on learning mix

There is a difference between a learning podcast and a podcast for learning. The first is a show about a subject, made for anyone interested, which people learn from as a side effect of enjoying it. The second is an instrument built into a designed learning experience, with objectives, an audience defined in advance, and a job to do.

They are made differently, resourced differently and measured differently. Most of the disappointment I see comes from organisations that commissioned the first and expected the second.

6. What it costs

Recently an academic at a VET institution told me that they had tried podcasting for learning but stopped because they found it too expensive. My jaw dropped. Perhaps this is because I run four podcasts, three of them completely on my own, from home, with no support other than commercial sponsorship, and my running costs are, if not insignificant for me, low enough to seem like a rounding error for most organisations – and yet I receive regular plaudits from listeners and other podcasters (including most recently JD Dillon, ex-Disney) for their quality.

This comment made me think deeply, though, about the real cost of podcasting for people who don’t have my particular abilities and resources, my background in creative arts, and freedom to experiment.

I will make a broad statement at the outset: podcasts are relatively cheap to produce compared to almost any other media. All you need to get started are a microphone, a computer and a hosting account. The technical bar is low. That’s why there are so many of them. They’re cheaper to produce and maintain than video (by a factor of ten, in my experience). Even when you add video to an audio-first podcast, audience expectations of production quality tend to be set by YouTube rather than, say, a Ken Burns documentary. And costs have even come down in the seven years I have been podcasting, due to AI.

Pre-edit audio treatment that used to take me an afternoon now takes ten minutes. Research, question drafting, transcript work, soundbite selection, show notes, captions, B-roll research and social copy are all partly automatable, and I (partly) automate all of them.

What has not got cheaper or easier is the part that matters. A presenter who can hold a conversation. Editorial judgement about what is worth broadcasting and how it is presented. Access to guests worth hearing. And somebody whose actual job this is. Those are the real budget line, and they are the ones organisations sometimes underestimate, because they look like they can be absorbed by someone’s existing role. They can’t, not always.

On generating the whole thing with AI

You can now produce a listenable show with very little human involvement. Google’s NotebookLM has proved that. The question is really whether your learners will want to listen to it. Sure, you can do it: the question is, should you?

Listeners have views on the matter. Edison’s Podcast Consumer 2026 found that 62% of weekly podcast consumers consider generative AI a threat to the credibility of podcasts, and 59% see it as a threat to creativity. The same audience is comfortable with AI in the back office – 69% approve of it for research and 69% for brainstorming ideas. The objection is to synthetic content, not to efficient production.

That maps onto the evidence in section 2. Trust is the property that makes the medium useful for learning, and trust attaches to a person. Automate the person and you might have removed the very thing your learners came for. And the academic literature we’ve drawn on in section 5 would seem to argue that the training types most suited to podcasting are also those that benefit most from distinctively human qualities.

There is an exception worth naming. For low-stakes purposes – a spoken summary of a document, an audio version of a briefing nobody was going to read anyway – an automatically generated feed may be exactly enough, and pretending otherwise would be dishonest. Just be clear exactly what it is you are making, and for what purpose.

A note on AI

I have current, detailed advice on running a podcast workflow with AI, and give keynotes and practical workshops about this subject regularly. However, my personal view is that, in a field that changes so fast and so much, any printed advice will tend to keep like fish. I therefore choose to deal with it in talks and directly with clients rather than in a document like this.

7. If you’re doing it already and it isn’t working

Some triggers for troubleshooting.

Nobody listens. The internal show with an audience no larger than the team that makes it. Usually a distribution and permission problem rather than a content one: people were probably never given time to listen, or told that it counted as work.

It stops at episode five. Of all the podcasts ever launched, about 11% are currently active. That’s one in nine. Enthusiasm alone will get you three episodes. Good planning, talent and a commitment to getting real value for learners could get you to thirty and beyond. (Once you get past 200, as in my case, it’s just a habit you would find it hard to break!)

The monologue. A senior person talking uninterrupted for forty minutes. Seniority isn’t necessarily the best qualification for the job – and college graduates who have been subjected to three years of death by lecture will know that the individuals who can make this format sing are actually quite few and far between. Interviews generally work better in getting that knowledge out.

Downloads as the measure. Downloads tell you about your promotion. They tell you nothing about whether anyone finished it, understood it or did anything differently as a result. Fifty of the right listeners can be a success and five thousand of the wrong ones a failure.

No editorial line. A show that cannot say anything that has not been signed off by three departments will tend to be bland, and blandness turns people off.

No owner. Added to somebody’s job on top of their job. It is the first thing dropped when the quarter gets busy. Something I quite often encounter with clients is a fuzziness about who in the team is responsible for actually getting the thing out. As an external consultant it can be hard to insist on this, so it’s worth stating up front: this is a common cause of failure.

Slop. Listen Notes removed 7,511 AI-generated fake podcasts from its database in February 2026 and 5,942 in January, both monthly records. Whatever else is happening in this medium, the volume of automatically generated audio is rising fast, and audiences are getting better at recognising it. I have had conversations with vendors who are selling completely AI-generated podcasts to L&D – and also to at least one Head of Learning who was enthusiastic about the idea. But what I don’t see is much mainstream uptake. Meanwhile, the AI backlash is building …

8. Ten questions

Be honest. The answers should tell you whether podcasting belongs in your mix or not. (No is a useful outcome!)

1. What should someone be able to do differently after listening? A weak answer names a topic area rather than a change.

2. Is there anything in your mix that already does this, and if so, does it do it well? If something already does it, you are proposing a replacement, not an addition.

3. Who is the audience, and what else are they doing during the time they would spend listening? If you haven’t thought about the context of use, you don’t have a distribution plan.

4. Is the material ‘transient-friendly’? Ideas, judgement, argument and stories, yes. Procedures, figures and anything that must be got exactly right, no.

5. Are you choosing audio for a reason about the content, or just because it is fashionable with the top team? “The exec team liked a podcast they heard” is a real reason. It is not a content reason.

6. Who is/are the host(s), and will people want to spend time with them? Seniority is not the qualification. Curiosity is.

7. Can you get guests/experts worth hearing, repeatedly? One good guest is an episode. A pipeline is a show.

8. Whose job is it to manage this, and what has to come off their plate to make room? If the answer to the second half is ‘nothing’, you have your prediction.

9. What comes out of the budget to pay for it, and is that trade-off worth what you might have to forgo? Answer this before you ask for the money, not after.

10. Can you sustain six to twelve episodes, and how will you know if it worked? Decide the measure now, while it is still cheap to be honest about it.

9. If the answer is yes

If you’ve read this paper and answered the questions above, you’ll probably have a much clearer idea about whether podcasting belongs in your mix and why. But there will probably still be questions about the where, what, when and who?

If you would like to talk these through, I offer a half hour with no charge and no deck. It is genuinely a conversation, and if my honest view is that a podcast is still the wrong answer for you, that is what you will get.

Book a half hour: https://calendly.com/johnhelmer/30min

The Learning Hack on Substack: https://learninghack.substack.com/archive

References

Statistical and industry sources were verified against the publishing organisation’s own material.

• Acast (2025). Podcast Pulse 2025. Global report published 4 November 2025. Research by Differentology: 2,500+ podcast listeners across nine markets (USA, UK, Australia, Canada, France, Germany, Mexico, Norway, Sweden). Acast is a podcast advertising company.
• Anderson, L. W. & Krathwohl, D. R. (eds) (2001). A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom’s Taxonomy of Educational Objectives. Longman.
• Baddeley, A. D. & Hitch, G. J. (1974). Working memory. In G. H. Bower (ed.), The Psychology of Learning and Motivation, vol. 8. Academic Press.
• Baddeley, A. D. & Hitch, G. J. (1977). Recency re-examined. In S. Dornic (ed.), Attention and Performance VI. Erlbaum, pp. 647–667. (Cited here only to correct a common misattribution.)
• Biggs, J. & Collis, K. (1982). Evaluating the Quality of Learning: The SOLO Taxonomy. Academic Press.
• Biggs, J., Tang, C. & Kennedy, G. (2022). Teaching for Quality Learning at University, 5th edition. Open University Press / McGraw-Hill.
• Cahill, L. & McGaugh, J. L. (1994). β-Adrenergic activation and memory for emotional events. Nature, 371, 702–704.
• Carvalho, A. A. et al. (2009). Taxonomy of podcasts in higher education, including design recommendations on length. Cited here as reported in Thomas et al. (2025); I have not read the original.
• Cedefop / European Union. The European Qualifications Framework: level descriptors. Eight levels defined as knowledge, skills, and responsibility and autonomy.
• Chater, N. (2018). The Mind Is Flat. Allen Lane.
• Dolcos, F. & Denkova, E. (2014). Current emotion research in cognitive neuroscience: linking enhancing and impairing effects of emotion on memory. Emotion Review, 6(4), 362–375.
• Edison Research at SSRS (2026). The Infinite Dial 2026. Published 12 March 2026. National survey, January 2026, n=2,050, US age 12+, approximately half drawn from a probability-recruited panel, weighted to census.
• Edison Research at SSRS (2026). The Podcast Consumer 2026. Published 4 June 2026, drawing on Edison Podcast Metrics and Share of Ear.
• Emergence Capital (2018). Estimate of the deskless proportion of the global workforce. Methodology not published.
• Gagné, R. M. (1985). The Conditions of Learning, 4th edition. Holt, Rinehart & Winston. First edition 1965.
• Guile, D. & Griffiths, T. (2001). Learning through work experience. Journal of Education and Work, 14(1), 113–131.
• Hall, E. T. (1963). A system for the notation of proxemic behavior. American Anthropologist, 65(5), 1003–1026.
• Hall, E. T. (1966). The Hidden Dimension. Doubleday.
• Harrison, J., Archer-Boyd, A. W., Francombe, J., Pike, C. & Murphy, D. T. (2023). The relationship between environmental context and attentional engagement in podcast listening experiences. Frontiers in Psychology, 13:1074320. University of York AudioLab with BBC Research and Development; online questionnaire, 264 listeners.
• Heilesen, S. B. (2010). What is the academic efficacy of podcasting? Computers & Education, 55(3), 1063–1068.
• Horton, D. & Wohl, R. R. (1956). Mass communication and para-social interaction: observations on intimacy at a distance. Psychiatry, 19(3), 215–229.
• Lave, J. & Wenger, E. (1991). Situated Learning: Legitimate Peripheral Participation. Cambridge University Press.
• Leahy, W. & Sweller, J. (2011). Cognitive load theory, modality of presentation and the transient information effect. Applied Cognitive Psychology, 25(6), 943–951.
• Listen Notes and Podcast Index. Active show counts and removals of AI-generated podcasts, reported March 2026.
• Mayer, R. E. (2021). Multimedia Learning, 3rd edition. Cambridge University Press.
• McGarr, O. (2009). A review of podcasting in higher education: its influence on the traditional lecture. Australasian Journal of Educational Technology, 25(3), 309–321.
• Meyer, J. H. F. & Land, R. (2005). Threshold concepts and troublesome knowledge (2): epistemological considerations and a conceptual framework for teaching and learning. Higher Education, 49, 373–388.
• Ofcom (2026). Media Nations 2026, published 29 July 2026, reporting RAJAR data.
• Shackleton-Jones, N. (2026). How People Learn: A New Model of Learning and Cognition, 3rd edition. Kogan Page.
• Singh, A.-M., Marcus, N. & Ayres, P. (2012). The transient information effect: investigating the impact of segmentation on spoken and written text. Applied Cognitive Psychology, 26(6), 848–853.
• Sweller, J., Ayres, P. & Kalyuga, S. (2011). Cognitive Load Theory. Springer.
• TalentLMS (2026). L&D Report: The State of Workplace Learning. Vendor-published.
• Thomas, M. A. M., Boivin, K., Brehm, W., Aktas, F., Karaagacli, B. C., Cho, I. & Kyei Mensah, P. (2025). Podcasts as pedagogy in higher education: a scoping review to map and advance the field. British Journal of Educational Technology. DOI 10.1111/bjet.70021. Open access; 91 publications reviewed.