This Week

The Current

Issue 11 19 October 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 the law that would let Australians switch your algorithm off went into Parliament, Meta is adding a shopping step to ads you already have running unless you stop it before the 28th, and Instagram shipped the cheapest idea for a post you'll get all month.

Worth Knowing

Two basics underneath all of this

A trending sound can carry a post further than the post deserves

Audio is one of the few free levers left, because when a sound is moving on a platform the app has a reason to keep showing clips that use it, and that reason applies to your clip too even when what you've filmed has nothing to do with whatever the sound is famous for. So a tiler showing a finished bathroom over a sound that's currently everywhere will usually travel further than the same clip with the radio on in the background, and it costs you nothing but picking from the audio list instead of ignoring it. I'd keep two limits in mind, though: the lift is real but it's temporary, since a sound that's peaked is just another sound, and it never rescues a clip whose first second gives nobody a reason to stay. Use it as a tailwind on something you'd have posted anyway, not as the idea itself.

One big post won't build the business, the boring run of them will

Nearly everyone who comes to us having tried social media on their own has had one post do unusually well, and what they remember is the week it happened rather than the six months either side of it, so the plan becomes trying to make that happen again. I think that's the single most expensive mistake in this whole area, because a spike brings you strangers who watched one thing, while a steady run of posts brings you the much smaller number of people who've now seen you four or five times and have started to think of you as the person who does that thing locally. The second group is the one that books. So if you've got twenty minutes a week to spend on this, spend it the same way every week rather than saving it up for something clever, and judge yourself on whether you turned up rather than on whether anything took off.

Australia's feed law is in Parliament, and the part that matters to you is open for comment

On Thursday 15 October the Communications Minister, Anika Wells, introduced the Online Safety Amendment (Digital Duty of Care) Bill 2026 into the House of Representatives. We ran the draft version of this as the lead back in Issue 06, when it was a consultation paper with no date attached, and the thing that's changed is that it's now an actual bill in front of Parliament, with a Senate inquiry ahead of it and the government saying it wants this done before the end of the year.

The headline part, for anybody who posts for a living or for a business, is the measure the government calls My Feed, My Way. Under it, Australians over 16 would get to choose whether their main feed is the recommended one the platform builds for them, or a feed of just the accounts and pages they've chosen to follow. The rest of the bill is a general duty on platforms to provide a safe environment, minimum safety standards that reach beyond social media to online games, apps and AI chatbots, specific protections for under-18s, and penalties up to $109.2 million for failing the duty. Two things changed after the consultation round: the ministerial power to add new categories of harm by rule has been taken out, so Parliament would have to legislate any new ones, and professional news and entertainment media are exempt.

Reaching people who don't follow you has always been the platform's decision to make, and this would hand a version of it to the person scrolling.

The reason I'd put this at the top rather than treat it as politics is that the whole mechanism small accounts grow by is recommendation to strangers. When a reel of yours does well, almost none of that reach is your followers, it's people the app decided to show you to, and that decision has always sat entirely with the platform. This bill would let a person opt out of being shown you. Nobody knows how many will, and I want to be plain that nobody does, including the people who drafted it, because most people leave defaults where they find them. But the direction is clear enough that it's worth understanding now rather than in whatever month it switches on.

It's also not law yet, and that distinction matters. Labor's majority means the lower house is close to a formality, but the Senate is a negotiation, and the minister has said she'll consider Greens proposals including letting individuals sue platforms over harm. Until it passes, the duties and the penalties are proposals.

The part you can actually do something about is separate from the bill. The rule that would make the follow-only feed a real requirement, the Online Safety (User Empowerment Tools) Determination 2026, is out as an exposure draft on the department's have-your-say page, under Choice of feeds on social media. When I checked, the page listed submissions closing on 5 November, but the listings I could see didn't all agree on the date, so check the page itself before you count on it. Worth saying honestly: one submission from one small business is not going to move a bill, and the department may publish what you send.

Do this week

Nothing to change in any account. If you want ten useful minutes out of it, open your Instagram or Facebook insights and find the split between followers and non-followers in your reach over the last ninety days, because that non-follower number is exactly what this bill is pointed at, and knowing it now is the only way you'll be able to tell later whether anything actually changed.

Meta is adding a shopping step to your running ads on 28 October

Around Tuesday 14 October advertisers started seeing a notice inside Ads Manager saying that from 28 October, Meta will add product browsing to their existing Facebook and Instagram ads unless they opt out first. Product browsing isn't new in itself, it's been sitting among the Advantage+ creative options as something you could switch on per ad. What changes on the 28th is the default for ads that are already running, which is why it's worth looking at rather than filing away.

What it does is put a step in between the tap and your website. Instead of going straight to wherever your ad points, someone who taps it may first see product information and personalised suggestions inside the app. There are two ways out, reported consistently: an opt out button in the notice itself, which only works before the 28th and appears to cover the existing ads on that account, and the per-ad toggle under Advantage+ creative and destination settings, which you can change at any time including afterwards, one ad at a time.

An extra step between the tap and your booking page is not a small thing when the whole point of the ad was the booking.

Whether this is good or bad for you depends almost entirely on what your ads are for. If you sell physical products out of a catalogue, an in-app browse is plausibly a help, and Meta's notice says that on average businesses saw 3% more conversions with in-app product browsing than campaigns without it, from an experiment it ran in August. I'd treat that as a sales line rather than a finding, because there's no published method behind it and it's Meta's own number about Meta's own feature. If your ads exist to get somebody onto a booking page, a quote form or the phone, then an extra surface between the tap and that page is the thing to think about, and nobody outside Meta can tell you what it does to your cost per lead.

The honest limits are worth stating, because they're unusual for a change this size. As of the middle of last week Meta had published no help centre article, no changelog entry and no newsroom post about switching this on for existing ads, so the date and the detail come from the advertiser notice and the trade coverage of it rather than from Meta directly. The notice also doesn't say what happens to brand new ads, duplicated ads, or ads created through the API, and no country list has been published, so whether Australian accounts are all in it is genuinely unknown.

Do this week

Open Ads Manager and look for the notice, and make the call before the 28th, because after that the one-click opt out is gone and you're doing it ad by ad. If you sell products from a catalogue, leaving it on is defensible. If your ads point at a form, a booking page or a phone number, I'd turn it off and note your cost per result for the fortnight either side so you've got something to compare.

Instagram will hold somebody else's Reel on screen while you film your answer

Reported through the week of 15 October by Social Media Today, Social Samosa and others: Instagram has added an option called Use as reference. You open a reel that's under 90 seconds, tap the three dot menu, and it keeps that reel playing in a small window while you film your own response. It's on iOS at the moment, with no date given for Android.

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

The hardest part of posting every week isn't the filming, it's deciding what to say, and answering somebody else hands you the subject for free.

I'm putting this in because of what it costs you, which is nothing, and what it solves, which is the actual problem most people have. Nearly nobody stops posting because filming got hard. They stop because they sat down on a Tuesday with no idea what to make. Reacting to something is the cheapest prompt there is: a plumber answering a popular DIY video that's giving people bad advice, a cafe responding to somebody's take on coffee prices, a physio watching a stretch go round and explaining why it doesn't do what people think. You're not inventing a subject, you're borrowing one that's already got people's attention, and the reference window just removes the fiddling of watching a thing, closing it, and trying to remember what you were going to say.

Two other things landed in the same batch. The Edits app got tappable audio links, so you can point viewers at the specific song or sound in a reel, plus insights on what's trending with your own audience, which are limited to accounts over 10,000 followers, and custom style kits for saving a look. Less welcome: the same coverage says Instagram is removing reels linking, the feature that let you chain clips into a series, on 28 October. If you've built anything that relies on it, that's a fortnight's notice.

What nobody has published is whether any of this lifts reach, and I'd rather say that than imply it. There's no data, from Instagram or anyone else, that a reference-filmed reply performs better than one you made from a standing start. The case for it is purely that it makes the thing easier to do, which for most people is the binding constraint anyway.

Do this week

Find one video in your own trade from the last week that picked up a lot of comments, open it on your phone, tap the three dots and see whether Use as reference is there. Film a thirty second answer in your own voice, no script. If the option isn't showing, you're probably on Android, and the old way of watching it first and then filming gets you the same post with a bit more faffing.

You can finally tell Google which part of your budget comes last

Google announced two new budget controls on Tuesday 14 October, in a blog post from Tal Akabas, who runs product management for Google Ads, and its help documentation puts the start of the rollout at 15 October, in the Google Ads interface and Search Ads 360 first with the API and Editor to follow. They cover Search, Shopping, Performance Max and Demand Gen campaigns.

The first is channel prioritisation. You mark a channel as high or low priority, and Google funds the high priority ones first, staying inside your target if you're on target CPA or target ROAS, and only then lets what's left flow to the low priority ones. The second is budget allocation, which puts a ceiling on the low priority channels as a percentage of the campaign's average daily budget. Google's own example: 5% on a $1,000 average daily budget means it aims to keep that channel at or under $50 a day. Budget allocation isn't available on Demand Gen.

Best effort means Google is aiming at your limit rather than promising it, which is still a great deal better than no limit at all.

The reason this is worth your attention if you're spending a few hundred dollars a month is that the oldest complaint in small-business search advertising is a Search campaign quietly spilling into the Display network or search partners and bringing back a pile of cheap clicks that never turn into anything. Until now the answer was usually to turn the whole thing off and hope. Now there's a dial. Just note the wording: Google says its tools will aim to keep spend within the limit on a best-effort basis, so it's a target rather than a hard cap, and you should still check what actually got spent rather than assuming the number held.

There's a dated thing attached to it too. Google says the older Display Expansion behaviour on Search campaigns goes away in February 2027, and its guidance is that if you use the Display network today, you should set it to low priority with a 50% cap by then to keep roughly what you've got. That's a long way off, but it's the sort of deadline that arrives having been ignored.

Do this week

If you run a Search campaign with Display Expansion or search partners switched on, go into its settings and look for the new channel priority controls. Set the Display side to low priority with a small cap, leave it a fortnight, then compare your cost per conversion against the fortnight before. If nothing gets worse, that spend was doing nothing for you and you've just got it back.

Google's new best model is out, and it isn't the one in your free app

Google made Gemini 4 Argon generally available on Tuesday 13 October, after a limited release at the end of September to trusted testers. It's Google's new flagship, and it reaches people in three ways: through the API for developers, and in the Gemini app and in AI Mode in Search for people on the paid Google AI Pro and Ultra plans. It is not on the free tier, in the app or in the API.

On where it sits, the independent tracker Artificial Analysis says this puts Google back among the top three labs, with Argon roughly level with OpenAI's GPT-6 Astra and behind Anthropic's Claude Opus 5.5. We've declined to quote a specific score from that index for five weeks running and I'm going to keep declining, because the write-ups we can see still disagree on the exact figures, and a number that moves depending on who's reporting it isn't worth putting in front of you as though it were settled.

The practical read for somebody using one of these to draft captions or plan a month of content is less exciting than the headlines and more useful. The three paid assistants are now close enough together that which one you pick barely matters for this kind of work, and the thing that actually makes a difference is whether you're on a paid tier at all. If you've been testing AI on a free account and finding the writing flat and a bit generic, that's worth knowing, because the model people are describing this week is not the model answering you.

The only other detail worth flagging is pricing, and it's mostly somebody else's problem: the introductory API rate runs to 30 November and roughly doubles after that, which matters to you only if a tool you pay for is built on top of it and decides to pass that on. And the usual caution applies to all of this. Benchmarks measure reasoning on set tasks, not whether something sounds like you talking about your own business, which is the only test that matters here.

Do this week

Nothing to buy and nothing to migrate. If you already pay for one of the big three assistants, you've got a current model and there's no action. If you're on a free tier and you've written AI off for writing, try the same prompt on a paid account for one month before you decide, because you may have been judging a smaller model the whole time.

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

The top of the table actually moved this week, for the first time in a while. Google made Gemini 4 Argon generally available on Tuesday 13 October, and Meta put Muse Wonder out through its own API on Monday 12 October, labelled a preview. Artificial Analysis has Google back in the top three labs with Argon about level with OpenAI's GPT-6 Astra and behind Anthropic's Claude Opus 5.5. Sixth week running we're not quoting a score, same reason as before: the write-ups disagree on the figures. For captions and planning, any of the paid tiers is fine.

Best for images

Still GPT Image 2.5, and that's a repeat for the sixth week, which we'd rather say than dress up. Nothing landed this week that changes it. Google's Nano Banana 2.1, which we covered last issue, is still the cheap option if you're generating a lot and editing real photos rather than inventing scenes, and it's the one we'd point a small budget at.

Best for video

Nothing we can confirm changed, which is the sixth week in a row, and we're still not naming a leader because the public leaderboards we can see disagree with each other. If you need video this month, film it.

Worth flagging

The gap between the model in the headlines and the model actually answering you is getting wider. Google's new flagship isn't on any free tier, in the app or the API, and the same has been true of the best OpenAI and Anthropic models for a while. So if your opinion of AI writing was formed on a free account, it was formed on a smaller model, and that's worth knowing before you either write the whole category off or pay for it.

Nobody can tell you what take-up will be

The lead this week is a bill, and bills are the hardest thing to write about honestly, because the useful question isn't what the law says, it's how many people will actually use the switch it creates, and nobody knows that, including the people who drafted it. My own guess is that take-up will be small, since most people leave defaults exactly where they found them, but a guess is all that is, and if I'm wrong then the way a small account reaches strangers changes shape here before it changes anywhere else. What we're doing about it in the meantime is the only thing that seems sensible, which is paying more attention to the followers a client already has, because a follow-only feed is the one place a follower is worth precisely what everybody has always assumed a follower was worth. Our own numbers on that are thin, we're a young studio and the case studies are still filling out, so treat it as a direction we're leaning rather than something we've proved. The rest of the issue is the usual shape: two defaults changing underneath people who didn't ask for it, and one small free feature that's genuinely worth ten minutes of your Tuesday.