This Week

The Current

Issue 13 2 November 2026 9 minute read Past editions
A laptop screen showing marketing analytics
Photo by ThisIsEngineering via Pexels.

If social media for your business feels like a moving target, this is for you. Plain, honest notes each week on what's actually working, from two people who run it for a living. This week is mostly about where the clicks go: Google has started running ads inside its AI answers in Australia and your existing campaigns are already in it, Reddit told investors Google is sending it a million fewer visits a week than it used to, and Pinterest's ads got cheaper in the same quarter nearly everyone else's got dearer.

Worth Knowing

Two things worth holding onto this week

Your followers aren't where the new customers come from

Follower count does one job properly, which is making you look like a real business in the two seconds somebody spends on your profile deciding whether to trust you, so it isn't useless. What it doesn't do is reach anyone. The feed on all of these platforms mostly shows your posts to people who don't follow you, chosen by what they've watched before rather than by who they've followed, so the people who eventually become customers nearly always see you well before they ever tap follow. I think that's worth holding onto this week in particular, because two of the stories below are about search sending fewer people to websites than it used to, and recommendation is the discovery engine you've still got. The practical version is just that every post should make sense to a stranger: the town in it, what you actually do, who it's for, because that stranger is most of your audience whether it feels like it or not.

A cheaper auction won't fix an offer nobody wanted

Paid ads are an amplifier, so they make whatever you already have louder, and that works beautifully when something is landing and painfully when it isn't. If a post got decent reach on its own and brought you enquiries, putting money behind it usually works. If nothing you post gets a response and the offer itself is the problem, then buying reach just means more people deciding no, faster, and you've paid for that. I'm saying it again because one of the stories below is about ad space getting genuinely cheaper on one platform, and cheap is good news, but cheap only helps if the thing you're putting in front of people is worth putting in front of them. The order matters more than the price: find the thing that works first, then pay to show it to more people.

Google has started running ads in its AI answers here, and your campaigns are already in it

On Wednesday 28 October, Google announced on its ads blog that ads are now running inside AI Mode in ten countries outside the United States, and Australia is in that first ten rather than being promised for later. The full list is Australia, Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines and Singapore, with Brazil, Mexico and the United Kingdom named for later in the quarter. The post came from Shashi Thakur, who runs Search ads at Google, and Engadget, Search Engine Roundtable and Android Authority all covered it within a day.

AI Mode, if you haven't poked at it, is the part of Google where you ask a question in a full sentence and get an answer written back at you instead of a page of blue links. Ads inside it are labelled Sponsored. The part that actually matters to you is how a campaign gets in there, because there's nothing to opt into: standard Search, Shopping and Performance Max campaigns automatically qualify for the AI Mode auction, and there's no way to target AI Mode on its own or to opt out of it separately from Search. So if you were running Google ads in Australia last week, you are running ads in AI Mode this week, and you didn't do anything to make that happen.

There's no switch for this one, which is unusual lately, and you're in it because you were already in Search.

Google also turned on something called multiturn ads in the United States, where an ad can appear in the answer to a follow-up question rather than only the first one you asked. That part isn't here.

The timing explains itself if you look at the same day's other news. Alphabet reported its third quarter results on 28 October too, and Search and other advertising grew about 12% to roughly 63 billion US dollars, which came in under what analysts expected and was part of the company's first miss on both revenue and profit in nearly four years. On the earnings call, chief business officer Philipp Schindler said one of the reasons growth slowed is that AI Mode monetises less well than ordinary search results while the number of questions people ask it keeps climbing. Google's stated fix for that, on the same call, is scaling ads inside AI Overviews and AI Mode and expanding them into more countries. So the blog post and the earnings call aren't two stories, they're one.

Google told investors the AI answers don't earn enough yet, and the plan for that is more ads in the AI answers.

On what it does to your results, the honest answer is that nobody outside Google knows yet. Google says early results are promising and hasn't published anything underneath that, and I haven't found anything saying whether you'll be able to see AI Mode performance split out in your own reports, so assume for now that it's mixed in with everything else. The one detail I would actually act on is Google's own figure that AI Mode questions run about three times longer than a traditional search. A person typing a whole sentence is telling you far more about what they want than someone typing two words, so this is the kind of traffic where plain descriptive copy and long, specific phrases do well, and where a tight little three word keyword list does badly.

I'd also keep this in proportion. It isn't a new campaign type, it isn't a price rise, and nothing in your account changed overnight. It's a new place your existing ads can appear, inside an answer rather than beside a list, and the thing worth watching over the next couple of months is whether your cost per enquiry moves while you haven't touched anything, because that's the only signal you'll get.

Do this week

Don't move bids or budgets over this, because you've got no data to move them on. Do two cheap things instead. Open your search terms report and read the long, sentence shaped queries people typed, since those are the ones most likely to be answered in AI Mode, and check your ad copy actually answers them in plain words. Then open AI Mode yourself and ask it three or four questions a real customer would ask about what you do, written the way they'd write them, and look at what comes back and who is advertising beside it. Twenty minutes, nothing spent.

Reddit says Google is sending it a million fewer visits a week

Reddit reported its third quarter on Thursday 29 October, and the advertising half of it was strong: advertising revenue up 54% year on year to about 847 million US dollars, with total revenue around 895 million. The number everybody wrote about was the other one. Chief executive Steve Huffman said search referrals fell meaningfully through the quarter, and that during October they were down by more than a million referrals per week on average, and he put that down to Google's AI Mode and AI Overviews. CNBC and Search Engine Journal both covered it the same day.

Reddit hasn't published the baseline those million a week losses are measured against, so the percentage is genuinely unknown and I'm not going to turn it into a scarier looking share than the company itself gave. What Reddit did say is that the loss fell hardest on logged out visitors arriving from the web, that it expects the decline to continue into the fourth quarter, that daily users could fall by a low double digit percentage from a base of around 130 million if it keeps going, and that it's testing making people log in, while spending more on marketing to replace the reach it's losing.

A company with every reason to sound calm about Google chose to put a number on it instead, which tells you something.

Now, why a tech company's earnings call is in a small business newsletter. Reddit isn't your competitor and its traffic problem isn't literally your traffic problem, but this is the clearest dated evidence anybody has published lately that the clicks coming out of Google search are shrinking, and the pipe that carries somebody from a Google result to reddit.com is the same pipe that carries somebody from a Google result to your website. Reddit can see the trend because it's enormous and it has to tell the market every three months. A plumber in Bayswater with four hundred sessions a month cannot see a trend like that in their own analytics, because the month to month noise is bigger than the signal, so a disclosure like this is about the only way to know whether a quiet few months is something you did or just the weather changing.

Read it next to the story above and you get both halves of the same week. Ads are going into the AI answer, and the clicks coming out of it are falling. The reasonable reading, I think, is that Google search is turning into a place where people get the answer rather than a place that hands them on to somebody, so if your plan for being found rests on a person clicking through from a search result to your website, that plan is slowly getting worse through no fault of yours.

What I'd avoid is the panic response. Don't rebuild your website over this, and please don't buy anything from anybody promising to get your business into AI answers, because there is no reliable published method for that yet and the people selling it know that better than you do. What is defensible is boring. Your Google Business Profile, which shows up in Maps and in local results and can make the phone ring without anybody visiting a website at all. Your social accounts, where being found is recommendation rather than search, which is exactly the point of the first Worth Knowing note above. And your own list of past customers, which is the only audience in this whole business that nobody else can reprice, rerank or reroute.

Do this week

Open whatever analytics you have, set it to the last twelve months, and look at organic search visits as a line rather than a number. If it's sliding, you now know it probably isn't something you broke. Then spend the next half hour on the channel that doesn't route through a search click at all: check your Google Business Profile has the right hours, the right phone number and photos from this year, because for a lot of local businesses that listing has quietly become the visit rather than the ad for it.

Pinterest's ads got cheaper while nearly everyone else's got dearer

Pinterest reported on Thursday 29 October as well. Revenue was 1.17 billion US dollars, up 12%, and monthly users hit a record 661 million, up 10%. Underneath those, the two numbers that matter if you buy ads: ad impressions grew 17% year on year, and the average price of an ad fell 4%. On the earnings call the company explained the price drop plainly enough, which is that impressions grew faster than revenue did.

Ad views
+17%
Revenue
+12%
Users
+10%

Pinterest's year on year growth for the third quarter of 2026, from its own results release of 29 October 2026 and its earnings call. "Ad views" is ad impressions. Ad space grew faster than the money spent buying it, and that gap is why the average price per ad fell 4% over the same three months. The 4% fall itself isn't drawn here, because these bars only run upwards from zero.

Put that beside last week's issue and you get a real contrast. In the same three months, Meta's average price per ad was up 14% worldwide, and Pinterest's was down 4%. It isn't that Pinterest is cleverer, it's supply: Pinterest created ad space faster than advertisers turned up to fill it, and when that happens the auction gets cheaper for whoever is standing there.

Cheap ad space isn't a strategy, it's just a stretch of time when a mistake costs you less to make.

Four honest limits before anybody moves money. First, that's a global average across every Pinterest advertiser on earth, so it is not a quote for your account, and what you pay is still set by your audience, your category and the time of year. Second, Pinterest doesn't publish a country breakdown of those 661 million users, it reports by region, so I can't tell you how many of them are in Australia, and a cheap auction is worth nothing at all if your customers aren't in it. Third, Pinterest missed both its own revenue guidance and the analyst estimate, and management blamed cross border sellers pulling spending after Europe changed its rules on low value imports, so part of what made the auction cheap is advertisers leaving, which is not the same thing as a healthy place to be. Fourth, prices move, and the last quarter of the year is the dearest one nearly everywhere, so a number from July to September isn't a promise about December.

The more useful question is whether it's your platform at all, because Pinterest behaves differently to the others. People use it to plan rather than to scroll, so it does well for food, homes and renovations, weddings and events, hair and beauty, clothes, gardens, anything somebody saves now because they're doing it in three months. If that's roughly your business, a quarter where space is 4% cheaper and 17% more plentiful is a reasonable quarter to run a small test. If you sell something people need urgently and locally, a locksmith, a physio, a tow truck, then this isn't your platform and a discount on the wrong room is not a bargain.

Do this week

Only if yours is the planning kind of business: take a hundred dollars you'd have put into Meta this month and run it on Pinterest instead, using the one organic image that already performed best for you, pointed at the page on your site that actually converts rather than the home page. Give it a fortnight, then compare cost per enquiry against what the same hundred dollars does on Meta. You're not buying sales at that size, you're buying an answer.

Instagram is testing 90 minute videos, and they're sideways

On Monday 26 October, Adam Mosseri and Tessa Lyons announced in a video that Instagram is testing a new format called Features: long form, horizontal video, up to 90 minutes while it's being tested. It started with creators in Brazil and opened to a selected group of creators in the United States from the 26th. TechCrunch, Social Media Today and 9to5Google all covered it, and a long tail of other outlets followed.

A person checking a social media grid on their phone next to a laptop
Photo by Plann via Pexels.

The horizontal part is the genuinely surprising bit. Instagram has spent five years training every one of us to hold the phone upright, and Mosseri himself has argued publicly against long form video on the app before, on the grounds that short videos get shared and talked about between friends far more. He now describes short and long form as symbiotic, so the position has moved, and that's worth noting even if the format never reaches you.

There's a bridge feature in it called Clips as Reels, where short vertical clips act as previews and anybody who taps gets taken through to the full horizontal version. One write-up says up to eight Reels can point at a single Feature, which I'd treat as reported rather than confirmed since I only found it in one place. If you read Issue 11 you'll notice the timing: Instagram removed Reels linking, the older feature that chained clips into a series, on 28 October. So inside one fortnight one way of linking clips together disappeared and a different one turned up.

Instagram spent five years teaching everybody to film vertically, and has just asked a few thousand people to turn the phone the other way.

What should you do about it this week? Nothing, and I want to be straight about that rather than dress a test up as an opportunity. It's a limited test with selected creators in two countries, there's no date for everybody else and no mention of Australia anywhere in it, and there's no revenue share for creators yet either, with Meta's own line to TechCrunch being that it's too early to share a long term monetisation model. Reports even disagree about where these videos appear, Explore for certain and the main feed according to some, which tells you how early this is.

It's still worth two minutes of your attention, for one reason. Every one of these platforms is converging on carrying every format at once: YouTube went after short vertical video, and Instagram is now going after long horizontal video, and the old advice to shoot natively for the platform is slowly becoming advice to shoot natively for the format instead. The practical read for a small business is that if you've ever filmed something long and genuinely good, a walkthrough of how you make the thing, a proper how-to, a conversation with a supplier, and then had nowhere on Instagram to put it, that gap is what this is aimed at. Don't start making long video for a format you can't post to yet. Just stop deleting the long footage.

Do this week

Nothing to switch on, nothing to sign up for. If you do have long footage sitting on a hard drive, the place it can earn something today is YouTube, which has wanted horizontal video the entire time and has search sitting behind it, so a twelve minute honest explanation of how you do your job can still be found there in two years. Upload one. If Features ever reaches Australia, you'll already have the thing it wants.

What's actually good right now

Disclosed upfront: this newsletter is partly written with Claude, so take our AI opinions here as disclosed rather than neutral.

Best all-round assistant

Nothing changed at the top for writing and planning, so whatever you used last week is still the right answer. The week's release was GPT-6.1 Terra on Monday 26 October, OpenAI's computer use model, which reads what's on the screen and takes actions across apps rather than just writing text back at you. It's in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users and in the API, and OpenAI says it costs about half what the previous version did to run. If you use an assistant for captions, planning and emails, that isn't an upgrade to anything you do.

Best for images

Still GPT Image 2.5, which is now a repeat for the eighth week running and we'd rather say that than pretend something moved. The public image boards we can see still have it on top. Google's Nano Banana 2.1 is still the cheaper pick if you're editing real photos rather than inventing scenes, and for a small business that's usually the more useful job anyway.

Best for video

We still can't name a leader honestly, which is the eighth week in a row, and the reason hasn't changed: the public boards disagree with each other. One we looked at this week has Alibaba's Wan 3.0 at the top of text to video, another had Kling v3 there instead, and the image to video rankings are different again. So pick on what you actually need, a clean product shot or a moving scene or spoken audio, rather than on a ranking that flips depending on which page you open.

Worth flagging

The two halves of this week are the same fact seen from either end. Google told its investors that AI answers earn less per question than a page of links, and its plan for that is to put more ads in the AI answers, including here. Reddit told its investors that AI answers send it a million fewer visits a week. One company is monetising the new shape of search and the other is being repriced by it, and the lesson for a small business is the same either way: the ways people can find you that don't depend on somebody clicking a search result just became more valuable than they were a year ago.

The week the answer stopped being a link

We sat down to find five unrelated stories and two of them turned out to be one story told from either side of it, which doesn't happen often. Google and Reddit reported their quarters a day apart, and between them they described the same change: the answer is becoming the destination instead of the thing that sends you somewhere. Neither company said that in those words, they both just reported a number, and the numbers point the same direction. We've written a lot this year about defaults quietly changing inside ad accounts, and this one is bigger than a default, because it's the shape of how people find anything.

Two honest follow-ups we chased this week and couldn't finish. Meta's product browsing default, which we flagged in Issue 11, was meant to switch on inside already running Facebook and Instagram ads on 28 October, and as of tonight Meta still hasn't published a help page, a changelog entry or a newsroom post about it, so we can't tell you what actually happened on the day. If you missed the opt out button in the notice, the per ad toggle under Advantage+ creative is still there any time. And the ChatGPT ads setting from last week, the one that rewrites your copy and can change the destination page, went live somewhere between the 29th and the 31st depending which OpenAI page you read, which is as resolved as that ever got.

Same caveat we run every week: our own numbers are thin while the case studies build out, so where we quote a figure it's somebody else's and sourced, and where we're giving you an opinion we say so. If you've seen something in your own account this week that contradicts anything above, especially on Google ads in AI Mode, we'd genuinely like to hear about it, because right now nobody has much.