This Week

The Current

Issue 15 16 November 2026 13 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 people find you and who gets to decide: ChatGPT has started letting businesses claim a profile and see the clicks it sends them, Google is testing a local result with nothing on it to tap, four platforms spent two days in front of a Senate committee arguing against Australia's feed law, and a new study says word of mouth is still how people find a small business.

Worth Knowing

Two things worth holding onto this week

What other people say about you is still the front door

We've made this point before from the angle of trust, that a review persuades somebody in a way your own caption never will, but one of the studies below puts it a different way that I think is more useful: it's not just more convincing, it's where people actually look first. Recommendations from friends and family and then customer reviews both came in ahead of social media posts as the way people choose a small business. So the practical version isn't "get more reviews" said in a nagging voice, it's that a lot of what you post on social is really doing its job when it gives somebody a reason to mention you to a friend, or a reason to go and leave a review, and if you've never once asked a happy customer to do either of those things then you've been doing the hard half and skipping the easy half.

Tell people exactly what to do next, inside the post itself

I'm bringing this one back early because of the Google story below, where the buttons that let somebody call you or get directions to you are quietly going missing from some search results. The lesson I take from that isn't about Google, it's that every button you rely on belongs to somebody else and can be moved or removed without anybody asking you. The part nobody can take away is the thing you actually say, so say it: call us on this number, book at this link, come in before five, reply to this post and we'll sort it. A clear plain ask converts better than a clever post that leaves people admiring you and then scrolling on, and it keeps working on whatever surface the post ends up on.

ChatGPT will now tell you how often it mentions your business

On Tuesday 10 November, OpenAI opened a beta called Businesses in ChatGPT, and the part of it that matters to you is a free dashboard called Business Hub. You claim your business, correct the details ChatGPT shows about you, and then see how often ChatGPT mentioned you and how many people clicked through to your website, over a 28 day window. Reuters reported the launch that day and SmartCompany covered the Australian end of it on 11 November.

I want to explain why this is the lead story in a week that also had a Senate inquiry in it. For the last couple of years, when somebody asks whether their business is showing up when people ask a chatbot for a recommendation, the honest answer has been that nobody can tell you. We said a version of that ourselves two weeks ago in Issue 13, when Reddit told investors that Google's AI answers were costing it more than a million search visits a week, and the point we made was that a small business can't see a shift like that in its own analytics because the noise is bigger than the signal. This is the first time anybody has handed this reader an actual number for it. It's a partial number from one chatbot, and I'll get to the limits, but it's a number.

The honest answer to "am I showing up in AI answers" has been that nobody can tell you. As of Tuesday that changed, for some businesses, a bit.

Now the limits, and they're big enough that a lot of you can stop reading here. Local stores and restaurants can only claim listings that already exist in the United States, so if your business is a shop, a cafe, a salon or a trade in Perth, there is most likely nothing for you to claim yet. The businesses that can act are online stores and online service businesses. Business Agents, which is the separate thing where customers chat to an agent that answers for your company, are United States only. And Australian people who went and tried it on 11 November reported being able to search for their listing and start a claim but then being put on a waitlist. OpenAI says it plans to expand to more markets and more business categories and has not given any dates.

Claiming is free and there's no ads account or enterprise plan involved. Verification runs off a business phone number, email address, website domain or a Shopify account, and then whoever manages the listing passes a separate identity check through a company called Persona. For an Australian business that means having your registered details to hand and matching, so the ABN and the registered business name rather than the trading name you use on your signage.

Two things worth understanding before you read anything into the numbers. A mention, for a local business, is counted when ChatGPT shows a result for that location, whether anybody opens it or not, so a mention is not a visit and not even really a read. And a click is a link click inside ChatGPT, which an OpenAI spokesperson told Digiday is not a completed visit or a purchase. Data also only appears once a business crosses some minimum activity threshold that isn't published, so an empty dashboard might mean nobody is asking about you, or it might mean not enough people are asking yet for OpenAI to show you anything.

The thing I'd most want you to take away is that claiming does not get you recommended more. Several write-ups say so plainly and it's worth repeating, because the moment a surface like this opens there will be people selling you placement in it. What claiming does is make sure the information is right, which matters because ChatGPT is assembling its description of you out of your own website, your reviews and licensed data from places like Yelp, and if any of that is five years old then so is the answer people are getting.

Do this week

If you sell online or you're a service business people buy from without walking in, go to Business Hub, search for your business, and start a claim with your ABN and registered name ready. Expect a waitlist. If you're a physical local business, don't spend your afternoon on it, because there's nothing for you yet. Do this instead, and it takes ten minutes: open ChatGPT and ask it the three things a customer would actually ask, so the best place for whatever you sell in your suburb, then your business by name, then a comparison with the competitor down the road. Write down everything it gets wrong. Then go and fix those things on your own website and your Google listing, because that's where it's reading them from.

Google is testing a local result with nowhere to tap

Search Engine Roundtable reported on Tuesday 10 November that Google is testing a version of the local pack inside its AI Overviews that drops the action buttons, and in some sightings the reviews, the photos, the address and the opening hours as well. Local search consultant Joy Hawkins found it on mobile, and in the version she saw the ordinary local pack was gone entirely and what replaced it was the business name and a few lines Google's AI had written about the business, and nothing else. No website, no call, no directions.

It's a test, and I want to be clear about that. Google hasn't confirmed any rollout, not everybody sees it, and plenty of things Google tests never ship. So this is a heads up rather than a thing that has happened to you.

The reason I think it's worth your ten minutes anyway is that it isn't a one off, it's a direction. Hawkins has been reporting for a while that call buttons have been disappearing from local packs generally, and she pointed at one study that found a call button on only about one in five map pack results. That study is small, a couple of hundred keywords, so I'd hold the exact share loosely, but the direction of it matches what the test looks like. And she made one more observation that I found harder to shrug off than the test itself: ads still get the buttons. Connecting a Google Business Profile to Google Ads with location assets can put your ad into the local three pack with a big call button that the free listings sitting next to it don't get.

The buttons on your listing were never yours. They belong to Google, and this is what it looks like when the owner starts moving them.

There was a second thing in the same week and the same corner of Google, which I'll flag because it's actually checkable. Search Engine Roundtable reported on Monday 9 November that Google appears to be adding Book Online buttons to some Business Profiles on its own, including on Barry Schwartz's own profile, with colleagues reporting the same elsewhere. In some cases the button goes somewhere broken and returns a server error. That's one trade report plus a first hand sighting rather than a confirmed change, so treat it as something to look at rather than something to be angry about, but a booking button on your listing that errors out is worth two minutes of your Monday.

For a lot of local businesses the Google listing is the single biggest source of phone calls, bigger than the website and bigger than social, and almost nobody knows that as a number because the calls just arrive. That's the actual exposure here. It isn't that Google is doing something to you this week, it's that a channel you depend on is one you don't control and can't see into, and the first time you'd notice a change is when the phone goes quieter and you assume it's the season.

Do this week

Open your Google Business Profile and do two things. First check the booking link, and if there's a Book Online button you didn't put there, tap it and see where it goes, and remove the link if it's broken. Second, look at the profile's own performance numbers for the last few months, the calls, the direction requests and the website clicks, and write them down. That's your baseline, and having it is the difference between noticing a change and inventing an explanation for one later.

Four platforms spent two days arguing against Australia's feed law

On Monday 9 and Tuesday 10 November the Senate Environment and Communications Legislation Committee held public hearings in Canberra on the Online Safety Amendment (Digital Duty of Care) Bill 2026, which is the bill we ran as Issue 11's lead, the one with My Feed, My Way inside it. Meta, Google, TikTok and Snap all gave evidence, and all four of them argued against it as drafted. Mi3 wrote the hearings up on 13 November.

Where the bill is up to: it passed the House of Representatives with amendments on 28 October, reported as 96 votes to 38, and it's now in the Senate. The committee is due to report by 20 November. The government wants it passed before the end of the year and needs crossbench support to get there, with the Coalition reported as opposing it. The duty itself is reported to commence after a twelve month transition period, which is the detail I'd hold onto, because it means nothing changes for you for a year even if this passes in the next fortnight.

The single most interesting thing any platform said, to me, was a number. Meta told the committee it runs more than 22 recommender systems, and asked how a ministerial determination would apply to individual features given there are that many of them. It wants clear thresholds, industry consultation and a common standard before a regulator steps in. Google warned the bill could end up requiring age checks on ordinary services like messaging, email and private file storage, which it put as possibly having to block under-18s from Gmail. TikTok's representative said the bill replaces certainty with ambiguity and subjectivity and warned platforms would over-moderate anything contentious rather than risk a penalty. Snap said it wasn't shirking its responsibilities. Senator David Pocock told the hearing that a Canberra girl had got past TikTok's age checks and was then shown posts promoting self harm. And the eSafety Commissioner, Julie Inman Grant, told the committee she has seen no sign of platforms over-censoring.

Meta runs more than 22 systems that decide who sees what, and not one of them is yours.

I keep coming back to the 22. Every one of those systems is a decision about who sees your post, and the whole reason a small account can grow at all is that some of them show your video to people who have never heard of you. My Feed, My Way would let an Australian over 16 choose a feed of only the accounts they follow instead. Nobody knows how many people would take that up, and I'm not going to pretend otherwise, but the shape of the risk hasn't changed since September: if enough people switch, reach to strangers gets harder, and reach to strangers is the part you can't buy your way out of cheaply.

One honest note on reading the hearings. Platforms argue against regulation, that's what platforms do, so "all four objected" is not evidence that the bill is bad. Some of what's being said about this bill online is also well outside what the mainstream coverage supports, including a framing that it's deliberately vague so it can be used for censorship, which the eSafety Commissioner directly rejected. I'd read the committee report on 20 November rather than anybody's summary of the vibe.

Do this week

Nothing, genuinely, and nobody can sell you preparation for a bill that hasn't passed. If you want to spend five minutes on it usefully, look at how much of your reach in the last month came from people who don't follow you, which Instagram and TikTok both show you under followers versus non-followers in your insights. If that share is most of your reach, you now know exactly which part of your growth a feed choice law would touch, and you'll know it a year before it matters.

Word of mouth is still how people find a small business

American Express published its 2026 Shop Small Impact Study on Thursday 12 November, ahead of Small Business Saturday in the United States on 28 November. The number I'd put in front of anybody who feels like they're failing at social media is the ranking of how people actually choose a small business: recommendations from friends and family came first at 65%, customer reviews second at 45%, and social media posts third at 39%.

Friends
65%
Reviews
45%
Social
39%

How consumers say they choose a small business, from the 2026 American Express Shop Small Impact Study, published 12 November 2026. Labels shortened to fit a phone screen: Friends is recommendations from friends and family, Reviews is customer ratings and reviews, Social is social media posts. Surveys run by Teneo for American Express: 1,008 United States adults in early to mid September 2026, and 503 owners or managers of non-franchise businesses with 2 to 99 employees between 11 and 21 September.

Two other figures from it. 88% of consumers said they're likely to shop at a small business this holiday season, which Amex says is the highest since 2012, and that's stated intent rather than money anybody has actually spent. And on the business side, 96% of owners said they're doing something new this year to get ready for the season, with 90% expecting to use AI tools somewhere in their marketing, operations or personalisation.

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

Social being third is not an argument for posting less, and I'd be annoyed if anybody used it that way. Look at what the top two have in common instead: both of them are somebody other than you doing the talking. A friend mentioning you over dinner and a stranger reading your reviews are the same mechanism, which is other people vouching for you, and the thing social media is genuinely good at is feeding that mechanism. Somebody has to see you often enough and like you enough to mention you, and somebody has to be asked at the right moment to write the review. That's what your posting is for, as much as it's for the sale itself.

The top two ways people choose a small business are both other people talking about you, which is also the thing your posting is actually for.

And the owner numbers, 96% changing something and 90% reaching for AI, are worth reading as a warning as much as a trend. If nearly every small business in a market is using the same tools to make the same kind of thing for the same season, then the thing that stands out is going to be whatever doesn't look like that. Which is, again, a real person in a real place saying something specific, filmed badly on a phone.

The caveats, properly. This is a United States sample, about a thousand consumers, which is small for a national read, and the fieldwork was done in September and published in November. The business half of it covers owners making sales at physical locations in retail, hospitality, personal services and arts or recreation, so it isn't every kind of small business and it's not an Australian one. And all of it is people telling a survey what they do rather than a record of what they did.

Do this week

Pick the five customers you've had the best interaction with in the last month and ask each of them for a Google review, by name, in person or in a message, with the link. That's it. Five is a number you'll actually finish, and reviews are the second strongest thing on that chart and the only one of the three you can directly go and get more of this week.

Google will now show you the clicks it decided not to charge you for

From Thursday 12 November, Google started rolling out an Invalid traffic card on the Overview page in Google Ads, under the Campaigns tab. Google announced it on its own ads blog and says it's rolling out globally, and Search Engine Land, Search Engine Journal, Search Engine Roundtable and PPC Land all covered it the same day.

What it shows is two dollar figures. The first is an estimate of what Google stopped you being charged, by filtering activity it judged invalid before it ever billed you. The second is the money it credited back for invalid activity it only spotted after the monthly billing cycle had closed. Click into View details and you get a daily trend chart, a campaign level table and a downloadable spreadsheet, and there's a new report in the Report editor under Performance Summary called Invalid Activity Credit that breaks the credits down by campaign and network, which PPC Land says is the first time Google has documented that report at all.

Invalid traffic, in plain terms, is clicks Google has decided weren't a real person with real intent, so bots, scrapers, accidental double clicks, and in the ugly cases a competitor or a disgruntled somebody clicking your ad on purpose. Google has always filtered it and has always credited some of it back, but it did that quietly inside your billing and most advertisers never saw any of it as a number.

Every small advertiser has wondered whether bots are eating the budget, and until Thursday nobody could answer it with anything but a feeling.

Google's own claim alongside this is that it has reduced advertiser exposure to invalid leads by an average of 30% so far in 2026. That's Google's figure about Google's own systems with no published method behind it, so I'd file it as a sales line rather than a finding, the same way we treated Meta's 3% conversions claim last month.

Four limits worth knowing before you go and look. It's an account level total, so it won't tell you where the bad traffic came from or which keyword attracted it. The way the savings figure is valued isn't published. The credits don't retrospectively change the Cost or the CPA in your main campaigns table, because those lock at the end of the month, so your reported cost per lead still includes money you later got back. And this is Google grading its own homework: the number is what Google's systems caught, not an estimate of what got through.

What it's actually good for, I think, is closing off a worry rather than solving a problem. If you've ever looked at a quiet month and wondered whether you were being clicked by bots, you can now look. And if the number is small, which for most small local accounts it probably will be, then that worry is answered and whatever's wrong with the campaign is something else, which is a genuinely useful thing to find out.

Do this week

If you run Google Ads, open the Overview page under Campaigns and find the Invalid traffic card. Write the two numbers down next to your monthly spend so you can see what share they are, then leave it alone. There's nothing to configure and nothing to opt into, and the one thing not to do is start making changes to a campaign on the strength of a number that's mostly telling you the filtering worked.

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 actually moved this week, after nine weeks of us saying it hadn't. Anthropic released Claude Fable 5.2 on Monday 9 November, which replaces Fable 5.1 as its most capable model anybody can use, with Anthropic saying output runs up to 30% faster. It came out alongside Claude Mythos 5.2, which you can ignore, since that one is restricted to vetted cybersecurity and life sciences organisations. Google also put out Gemini 4 Hydrogen on 12 November, which is the cheap fast one rather than its best, and the reason it's worth a line is that it's on the free tier and reportedly powering the AI answers in Google Search, which is the gap we flagged in Issue 11 when Google's actual best model landed on no free tier at all. That detail is from third party write-ups rather than Google's own post, so hold it lightly. Tenth week running we're not quoting you a single ranking score, for the same reason as always: the public boards disagree with each other.

Best for images

Same answer as last week, which is OpenAI's Images 2.6, released on 4 November and live in ChatGPT on every tier including the free one, on web, desktop and phone. The feature in it that a small business will actually feel is being able to save a reference image so a set of posts can keep the same look. Nothing confirmable changed this week: a couple of trackers list a new image model from MiniMax on 10 November, but we couldn't find the company's own announcement, so we're not naming it as a change. Google's Nano Banana 2.1 is still the cheaper pick if you're editing a real photo rather than inventing a scene.

Best for video

Ten weeks now with no leader we can name honestly, and this week there wasn't even a release to argue about: we found no confirmed video model launch in the window at all. The boards we can see still disagree with each other about who's on top, and the vote counts behind some of them are small enough that the order moves on noise. So the advice is the same as it has been since September. Pick on the job in front of you rather than on a ranking, and if you're choosing between the big ones, Google's Veo and Kling both do matched audio now, which is usually what decides it for a short ad.

Worth flagging

Hold this week's first two stories next to each other. OpenAI handed you a dashboard for a place you couldn't previously see into, and Google started testing a version of a place you've relied on for a decade with the buttons taken off. Those feel like opposite news and they're the same fact: the front door to your business keeps being rebuilt by companies you don't own and can't ring up. I don't think the answer to that is chasing whichever surface is new, because there'll be another one by autumn. The answer is noticing which parts actually belong to you, which is your website, your customer list and the people who already like you enough to mention you, and making sure none of those depend on a button somebody else can move.

The front door keeps getting moved

We don't pick a theme for these, we just read the week and write what's in it, and this week the week picked one for us. Four of the five stories are about the place where somebody first comes across your business, and in every one of them that place is owned and rearranged by somebody else. A chatbot starts keeping a profile of you. A search result loses the button that let people call you. A law about recommendation feeds goes through a Senate committee. And then a study quietly points out that the thing people still trust most is a mate telling them you're good, which is the one channel nobody has shipped an update to.

One thing we found this week and deliberately didn't run as a story, because it's dated to the week before and we missed it when it was new, but which you should know about anyway. Meta has a setting in Commerce Manager called Detected offers, which lets it take promotions it finds on your website or in your other ads and display them on your ads. PPC Land reported on 6 November, after digital marketing advisor Bram Van der Hallen flagged it, that a discount can carry into a copied ad even when promotions were switched off on the original, and Social Media Today covered it on 8 November. Meta's own help page says the way to stop crawled promotions is to turn Detected offers off in Commerce Manager, and that turning promotions off on a single ad won't do it. Nobody has established whether it's on by default, and the screenshots going round come from one account, so the only reliable version is to go and look at your own. Black Friday is on the 27th and an expired special is a bad thing to have running in a paid ad, which is the same point we made about Google doing this in Issue 10.

Two more we're carrying rather than burying. Meta's product browsing default from Issue 11, the one that was meant to switch on inside already running ads on 28 October, is into its fifth week with still no help centre article, changelog entry or newsroom post we can find, and the per ad opt out under Advantage+ creative is still there if you want it. And a list of Meta's November Ads Manager changes went round this week, several of them reportedly switched on in accounts without asking, including turning uploaded images into a carousel. We cut it because it traces back to one practitioner's thread and we couldn't find Meta's own list anywhere, so we're not writing it as fact, but it's worth half an hour in your own account's creative settings regardless. 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.