TRANSCRIPT 🔗
WEBVTT
00:00:00.000 --> 00:00:05.295
Welcome to Shallow DeepDive, where we take
a deep-ish dive into the shallow world of
00:00:05.375 --> 00:00:07.534
branding, advertising and creative.
00:00:07.614 --> 00:00:08.494
I'm Joe Qualls.
00:00:08.574 --> 00:00:13.869
Today, we're covering what a brand is
really buying when it pays to appear
00:00:13.949 --> 00:00:14.666
AI's answer.
00:00:14.746 --> 00:00:17.674
Here's a small moment that's becoming
ordinary.
00:00:17.754 --> 00:00:22.793
Someone asks the assistant a genuine
question — which accounting software,
00:00:22.873 --> 00:00:27.847
running shoe, somewhere to take the kids
these holidays — and gets back a calm,
00:00:27.927 --> 00:00:28.999
reasonable answer.
00:00:29.079 --> 00:00:33.862
And underneath it now, in a softly tinted
little box, sits a sponsored card.
00:00:33.942 --> 00:00:34.659
A brand.
00:00:34.739 --> 00:00:36.131
A logo, a line, a link.
00:00:36.211 --> 00:00:37.027
It's labelled.
00:00:37.107 --> 00:00:37.824
It's polite.
00:00:37.904 --> 00:00:41.919
It doesn't really look like an ad so much
as a helpful footnote.
00:00:41.999 --> 00:00:47.038
That box is what we want to talk about,
because it's one of the more interesting
00:00:47.118 --> 00:00:52.221
things to happen to advertising in a
while, and almost nobody is asking the
00:00:52.301 --> 00:00:53.373
question about it.
00:00:53.453 --> 00:00:54.845
Briefly, what happened.
00:00:54.925 --> 00:00:59.707
Earlier this year, OpenAI started putting
ads in ChatGPT — for free users, a
00:00:59.787 --> 00:01:04.954
sponsored card below the answer, with a
title, a short description, an image and a
00:01:05.034 --> 00:01:05.752
link.
00:01:05.832 --> 00:01:07.799
It's now live here in Australia.
00:01:07.879 --> 00:01:13.174
The company was careful, and the design is
careful: the box is labelled, it's tinted
00:01:13.254 --> 00:01:18.357
a slightly different colour, and the
promise attached to it is that the ads
00:01:18.437 --> 00:01:20.213
change the answer above them.
00:01:20.293 --> 00:01:21.812
The answer is the answer.
00:01:21.892 --> 00:01:23.028
The ad is separate.
00:01:23.108 --> 00:01:28.403
Whether or not you fully believe that last
part is, more or less, the whole episode.
00:01:28.483 --> 00:01:33.394
Advertisers believe in the surface, at
least: the program reportedly reached a
00:01:33.474 --> 00:01:37.553
hundred million dollars in annualised
revenue in about six weeks.
00:01:37.633 --> 00:01:42.927
That number tells you the demand was
already there, sitting behind the door,
00:01:43.007 --> 00:01:44.399
for someone to open it.
00:01:44.479 --> 00:01:46.831
The easy way to read this is "a new ad
00:01:46.911 --> 00:01:51.949
" Another place to buy attention, like a
search ad or a banner, only smarter and
00:01:52.029 --> 00:01:52.973
better targeted.
00:01:53.053 --> 00:01:58.220
And if that's all it is, then the only
questions are the usual ones — what does
00:01:58.300 --> 00:02:00.844
cost, who does it reach, does it convert.
00:02:00.924 --> 00:02:04.619
But that reading quietly misses what's
actually being sold.
00:02:04.699 --> 00:02:07.754
Because the thing for sale here is not
attention.
00:02:07.834 --> 00:02:10.313
Attention is cheap, and it's everywhere.
00:02:10.393 --> 00:02:12.745
What's for sale is proximity to trust.
00:02:12.825 --> 00:02:15.816
Think about why the placement is worth
anything.
00:02:15.896 --> 00:02:20.359
A person asked a question and got an
answer they've decided to believe.
00:02:20.439 --> 00:02:25.542
They're in a rare state for advertising —
open, unguarded, mid-decision, treating
00:02:25.622 --> 00:02:29.573
the thing on screen as a neutral helper
rather than a salesman.
00:02:29.653 --> 00:02:34.820
And your card arrives in that exact
moment, wearing a little of the answer's
00:02:34.900 --> 00:02:37.059
It doesn't feel like being sold to.
00:02:37.139 --> 00:02:38.787
It feels like being helped.
00:02:38.867 --> 00:02:41.411
That borrowed calm is the entire premium.
00:02:41.491 --> 00:02:44.930
You're not really renting the space next
to the answer.
00:02:45.010 --> 00:02:49.985
You're renting a bit of the answer's
credibility, for the length of one glance.
00:02:50.065 --> 00:02:55.167
And here's the part that's easy to miss,
because it works perfectly for the first
00:02:55.247 --> 00:02:57.535
advertiser and gets worse from there.
00:02:57.615 --> 00:03:02.590
The credibility you're borrowing isn't
yours, and it doesn't refill on its own.
00:03:02.670 --> 00:03:07.645
It belongs to the medium — to the general
belief that when you ask this thing a
00:03:07.725 --> 00:03:12.508
question, it tells you the truth as best
it can, with no thumb on the scale.
00:03:12.588 --> 00:03:15.963
Every sponsored card spends a tiny bit of
that belief.
00:03:16.043 --> 00:03:18.458
Not because any single ad is dishonest.
00:03:18.538 --> 00:03:23.193
Simply because the presence of money in
the room teaches people to wonder.
00:03:23.273 --> 00:03:28.440
And once someone has wondered even once —
was that recommendation the best one, or
00:03:28.520 --> 00:03:29.272
the paid one?
00:03:29.352 --> 00:03:31.255
— they don't just doubt the ad.
00:03:31.335 --> 00:03:36.310
They start to doubt the answer above it,
and the next answer, and the one after
00:03:36.390 --> 00:03:37.108
that.
00:03:37.188 --> 00:03:41.843
The reliability of every reply takes the
hit, not only the sponsored ones.
00:03:41.923 --> 00:03:46.706
So this is one of those situations
advertising keeps rediscovering and keeps
00:03:46.786 --> 00:03:47.921
managing to forget.
00:03:48.001 --> 00:03:53.296
The value of the placement rests on a
shared resource that no single advertiser
00:03:53.376 --> 00:03:55.344
and every advertiser draws down.
00:03:55.424 --> 00:04:00.527
The more successfully the whole industry
piles in, the faster the thing that made
00:04:00.607 --> 00:04:05.709
the placement worth buying — the sense
that the answer is neutral —
Load full transcript
gets used up.
00:04:05.789 --> 00:04:08.653
It's a lovely spot to be the first brand
into.
00:04:08.733 --> 00:04:11.468
It's a much worse spot to be the
thousandth.
00:04:11.548 --> 00:04:16.715
The card that felt like a helpful footnote
in February starts, eventually, to feel
00:04:16.795 --> 00:04:18.251
like exactly what it is.
00:04:18.331 --> 00:04:20.874
We're not here to tell you to boycott it.
00:04:20.954 --> 00:04:26.185
That's not our job, and there'll be good,
honest reasons to run these — a genuinely
00:04:26.265 --> 00:04:30.792
relevant product, in a genuinely relevant
moment, is a fine thing to be.
00:04:30.872 --> 00:04:35.911
But notice the question most brands are
actually asking, which is the small one:
00:04:35.991 --> 00:04:39.046
should we buy this, and what's it cost per
click.
00:04:39.126 --> 00:04:42.437
The better question is the one sitting
underneath it.
00:04:42.517 --> 00:04:47.812
What are we teaching people to feel about
our name when it turns up beside an answer
00:04:47.892 --> 00:04:48.644
they trusted?
00:04:48.724 --> 00:04:53.187
Because every placement is also a tiny
lesson in what your brand means.
00:04:53.267 --> 00:04:58.370
Show up as the useful, relevant thing in a
real moment, and you teach one lesson.
00:04:58.450 --> 00:05:03.553
Show up as the interruption in a moment
somebody wanted kept clean, and you teach
00:05:03.633 --> 00:05:04.896
another one entirely.
00:05:04.976 --> 00:05:10.079
The second lesson is expensive, and it
never appears in the click-through report.
00:05:10.159 --> 00:05:14.558
There's a deeper move here too, one we've
circled before on this show.
00:05:14.638 --> 00:05:17.501
The advertiser buys the box beside the
answer.
00:05:17.581 --> 00:05:19.869
But the real prize was never the box.
00:05:19.949 --> 00:05:21.405
It was being the answer.
00:05:21.485 --> 00:05:26.651
When someone asks "which one," and the
model — with no money changing hands —
00:05:26.731 --> 00:05:31.834
yours, because yours is genuinely the
clearest, best-documented,
00:05:31.914 --> 00:05:35.801
option in the category, you haven't rented
trust for a glance.
00:05:35.881 --> 00:05:38.873
You've earned the kind that doesn't get
used up.
00:05:38.953 --> 00:05:41.368
That's slower, and it's less glamorous.
00:05:41.448 --> 00:05:46.423
It looks like doing the actual work: being
legible to both people and machines,
00:05:46.503 --> 00:05:50.454
being the thing that stays true when
nobody's paying to say so.
00:05:50.534 --> 00:05:55.829
It's also the only position in this whole
arrangement that gets stronger as everyone
00:05:55.909 --> 00:05:57.045
else's gets weaker.
00:05:57.125 --> 00:06:02.164
So the sponsored card is worth watching —
not because it's sinister, it's mostly
00:06:02.244 --> 00:06:06.323
just clever — but because it runs on a
resource it can't replace.
00:06:06.403 --> 00:06:10.290
The first brands through the door get a
real, early advantage.
00:06:10.370 --> 00:06:12.209
The medium pays for it slowly.
00:06:12.289 --> 00:06:15.984
And so, in the end, does everyone who
piled in behind them.
00:06:16.064 --> 00:06:20.911
If you're deciding whether to be in that
box, then decide it, with open eyes.
00:06:20.991 --> 00:06:26.222
But decide the bigger thing first: whether
you're building a brand that has to rent
00:06:26.302 --> 00:06:31.405
trust by the click, or one that people —
and the machines that increasingly speak
00:06:31.485 --> 00:06:34.092
for them — already reach for on their own.
00:06:34.172 --> 00:06:37.612
Thank you for indulging us in today's
Shallow DeepDive.
00:06:37.692 --> 00:06:40.171
We hope you'll join us for the next one.
00:06:40.251 --> 00:06:41.579
Until then, take care.