Two thirds of searches now end without a click. AI search answered them without you.
Not because your show is bad. Because these systems read text, and most of what your show says was never written down anywhere they can reach.
SparkToro and Similarweb tracked US Google searches from January to April 2026. They found 68.01 percent ended with no click at all. Surfer SEO looked at 46 million AI Overview citations and found YouTube alone accounted for about 23 percent of them. The sources these systems pull from are fewer than most people assume, and getting into that pool is mostly a production decision.
Producing since 2014 for Salesforce, Honda & Stanford
The companies behind shows we have produced.
You paid for the research. The engine read it. Your competitor got named.
That is the whole problem in one sentence, and nobody says it out loud because it is embarrassing. Your team spends months producing something genuinely useful. An answer engine takes it in, builds a reply out of it, and recommends someone else by name. You funded the education of a machine that now sells for a competitor.
Ranking first on Google does not protect you. Ranking and being cited are two different jobs, decided by two different systems, drawing on two different source sets. Most brands are still only playing the first game and wondering why the pipeline went quiet.
Book a Strategy Call →January to April 2026, up from 60.45% in 2024.
AI Overviews now show on more than 20% of searches.
The single largest source in the set.
2026. Now live across nine markets. The unpaid window has a shelf life.
“We’ll look at AI search next year.”
Head of Marketing, whose competitor ChatGPT now recommends by name.
Four things decide whether a show is reachable. None of them are talent.
These are not predictions of citation volume. They are the properties that make a show retrievable at all, and every one of them is a production decision made before episode one.
The words have to exist as text
Retrieval systems index text. An episode without a public transcript is a title and a description, and everything said in the hour is invisible to them. This is the least glamorous item here and the one most often postponed.
Long enough to say something specific
The most-cited programs run long, typically an hour and often far more. A fifteen-minute episode leaves a thin transcript with little that is precise enough to quote back inside an answer.
Named people, not a house voice
Interviews with named executives, researchers and founders give a system entities it can resolve and attribute. A solo monologue gives it one. Attribution is most of what these engines are doing.
One subject, held long enough to mean something
Engines pair sources to subjects. A show that ranges across everything pairs to nothing in particular, which is why depth in one category beats breadth across several.
The podcast citation research is thin. We are not going to pretend otherwise.
There is a small industry of studies claiming to rank which podcasts AI engines cite. We read them. Most rank hosts rather than shows. Most are published by firms selling AI visibility services. At least one says in its own methodology that its scores are a directional estimate rather than a live measurement. We are not going to quote you a citation count that nobody can reproduce.
So this page uses a narrower evidence base and states it openly. The figures on the right are the ones that hold up when you click through, and they describe the environment rather than the podcast layer specifically. Everything else here is either a property of how retrieval works or a production decision we are recommending on the record.
Book a Strategy Call →Why the engines disagree about which podcasts to cite. And why chasing one is a waste.
People notice that ChatGPT and Perplexity recommend different shows for the same question and assume one of them is wrong. Neither is. Each was built to answer a different kind of question, and that shapes what it reaches for.
Built to answer research questions
Runs live retrieval on the query rather than working from a fixed training set, and favors sources it can quote and attribute cleanly. A published transcript is close to ideal input. Recency counts here more than almost anywhere else.
Weighted toward demonstrated expertise
Tends to surface credentialed voices and careful, technical material over popularity. A specialist show with a modest audience and real subject depth does better here than its download numbers would suggest.
Anchored to what people search
Generated against real query intent, so they track the question being asked more than the reputation of the source. This is also where YouTube's share is largest, which makes video the practical route in.
Skews to the widely known
Leans toward well-established business and entertainment names. It is the hardest surface for a new show to enter directly, and the one where an executive's guest appearance on a larger program does the most work.
Balanced across the open web
Sits between the two poles, weighing recognition and subject expertise together, and drawing across a wide base of indexed web content rather than a narrow authority set.
The dull characteristics, on every engine
Crawlable text, one clear subject, named people, a feed that keeps publishing. No engine penalizes any of these. That is the whole reason to build to them rather than to optimize for whichever engine is fashionable this quarter.
So just launch a podcast, right? That is where most of the money goes to die.
Most branded podcasts will never appear in an answer. Not because podcasting does not work. Because of six decisions made in the first month, usually by someone reasonable, for reasons that sounded sensible at the time.
“We’ll add transcripts later.”
Later never arrives. Until then the archive is audio, and audio is not what these systems read. Every episode published without one is a month of work that stays invisible to the layer you are trying to reach.
“Let’s cover whatever is trending.”
Engines pair sources to subjects. A show about everything gets paired to nothing, and a year of episodes ends up with no question it is the obvious answer to.
“Nobody has time for an hour.”
Short episodes feel considerate and leave almost nothing quotable behind. The programs that get pulled into answers are long, because length is where the specifics live.
“Our CEO can just talk for twenty minutes.”
A monologue gives a system one entity to work with. Long-form interviews with named, credentialed guests give it a network of them, all pointing back at you.
“Let’s do a ten-episode season and see.”
You will see nothing. Live-retrieval engines weight recency, and a feed that stopped six months ago reads as a closed archive while competitors keep publishing into the same subject.
“Downloads are up, so it is working.”
Downloads measure who pressed play. Citations measure what a machine could read afterwards. They are not the same currency, and only one of them ends up inside an answer somebody else is reading.
Launching a podcast will not make ChatGPT cite you.
We would rather lose the sale than pretend otherwise. Citation weight sits with large, established programs, and a new show does not join them by existing. Any agency telling you a launch gets you cited by Christmas is either not reading the research or hoping you have not.
So why build one? Because the mechanism and the ranking are different things. The ranking belongs to shows with a ten-year head start. The mechanism is crawlable text, subject focus, named people and a feed that keeps going. It works the same way on a show you own as on a show you appear on.
Build the first properly and two things follow. You accumulate a body of retrievable material in your own name, in your own subject. And your executives become worth booking on the programs that already carry the weight. That is a slower promise than the one you will hear elsewhere. It is the one the evidence actually supports.
Book a Strategy Call →Six production decisions. That is the entire playbook.
Resonate has produced podcasts since 2014: 3,000+ shows launched, 50,000+ episodes delivered. Below is how each property becomes something we actually do on your show.
Transcripts on your domain, indexed
Full text, structured with real headings, on a page you own. Not a summary, not trapped inside a player widget. The item on this page that does the most work, and the one most commonly skipped.
Learn moreThe subject, decided before episode one
We define the specific question set you intend to own, then hold every episode to it. A show about everything gets pulled into nothing.
Learn moreTwo surfaces, one recording
Audio and video from the same session. Surfer SEO's analysis of 46 million AI Overview citations put YouTube at roughly 23 percent of them, the largest single source in the set.
Learn moreA cadence that survives Q4
Weekly or every other week, held for years rather than for one enthusiastic quarter. A dedicated producer is the difference between a schedule and an intention.
Learn moreMore crawl paths to the same episode
Audiograms, videograms and transcripts put the same conversation on more surfaces, which means more routes by which a system can find it and more places it can be attributed from.
Learn moreDistribution that stays fresh
Perplexity retrieves live against the query with no knowledge cutoff, so recent material has a real advantage there. Publishing consistently is the strategy, not the admin around it.
Learn moreWe are recommending you build for AI. Just not that you hand it the edit.
This whole page argues that answer engines are worth building for. That is not the same as letting one cut your show. Below are three raw clips, each run twice: once through the automatic edit on a widely used AI editing tool, once through a Resonate engineer. Same source audio, same episode, no cherry-picking. In every clip you hear the AI version first, then ours. Headphones if you have them.
Where the cut lands
AI cuts on silence. A person cuts on meaning, which is why one version keeps the breath before the point and the other removes it.
What counts as a mistake
Not every “um” is a flaw and not every pause is dead air. Removing all of them is how a conversation starts sounding like a press release.
How two voices sit together
Levels, room tone, and the handoff between speakers. The part listeners never consciously notice until it is wrong.
The AI edit, first half of each clip. Fast, cheap, and genuinely useful for a rough pass. It does exactly what it is told and nothing it is not.
The Resonate edit, second half. Degree-trained engineers, Dolby Atmos and iZotope RX certified, three to five business day turnaround. Judgment, not just processing.
Not sure a show is the right instrument for you?
Thirty minutes covers which buyer questions you should be trying to own, what your current material looks like to a retrieval system, and whether podcasting actually fits. If it does not, we will say so and you will have lost half an hour.
Straight answers, including the ones that do not flatter us.
Where a question has an inconvenient answer, the inconvenient answer is the one below.
Get the Free Ebook →How many podcasts do AI engines actually cite with any regularity?
Why do different AI engines cite different podcasts?
What is a podcast discovery engine, and is AI search the same thing?
For a podcast network, how does being recommended by AI assistants compare with platform algorithm placement?
Why do transcripts matter so much for AI visibility?
Will launching a podcast get my brand cited by ChatGPT?
How long does building AI search visibility take?
Does video help?
Is it too late to start?
What is in the free ebook?
Take the research and run it yourself.
Nine pages, every figure traced to a named source, including a straight assessment of which sources in this field are worth trusting and which are not. Written to survive the colleague who checks your citations.
- What changed in AI search, with the sourcing behind each figure
- Which podcast citation studies hold up, and which do not
- The production specification of a retrievable show
- An honest read on what podcasting cannot do here
- A 90-day sequence you can run without hiring anyone
- A complete sources page
No credit card, no call booked, no drip sequence you cannot leave.
Looking for your sector specifically?
Resonate builds industry-specific podcast programs across regulated and consumer-facing sectors alike. Each page covers the compliance and positioning details particular to it.
Somebody is already the answer to your category’s questions.
It may as well be you.
Book thirty minutes, or take the research and do it without us. Both are genuinely fine.
No commitment. Thirty minutes. An honest read on whether this fits your situation.