One brand went from AI warning buyers off in 27 answers out of 42 to none, by handing the models the denominator
This week one brand stopped AI warning buyers off simply by telling the models how many happy customers sat behind its handful of complaints, and Google Lens turns out to be reading your photos rather than admiring them. Elsewhere, LLMs are doing word maths with your positioning, TikTok and YouTube are introducing students to universities more often than search engines do, and twelve real followers turn out to be worth more than 120,000 bots. Here are your five.
1. One brand went from AI warning buyers off in 27 answers out of 42 to none, by handing the models the denominator
AnswerShare put the same neutral reputation questions to the frontier models through the API, on behalf of a client with 75-plus positive reviews, five negative ones and two BBB complaints. Before: concerns raised in 27 of 42 answers, a warning to look elsewhere in all 27, a recommendation just 10 times, and only one answer that bothered to mention the company's size. What the models were missing was the bit that changes everything: 35,000 customers over 13 years, which turns seven complaints into a 99.98% satisfaction rate. So they published a sourced, warts-and-all llms-full.txt and served the same text from a CDN worker, which AI crawlers hit 744,566 times in two months against 154 visits to the file itself. Fourteen days later: zero warnings in 40 answers, 40 recommendations out of 40, and the complaints still mentioned every time, just with context attached.
Apply the full squint here: it's a sponsored post, the results are AnswerShare's own clients, the samples are around 40 prompts apiece, and the article needs a whole section titled ‘Is this cloaking?’, which tells you somebody has asked. The underlying point is sound and pleasingly old-fashioned, though. A model that has been told to be cautious will treat five bad reviews as the whole story if nobody tells it the rest, and the BBB complaints in question turned out to be about one salesperson's tactics, not the actual service. Seven complaints is either a scandal or a rounding error, depending entirely on what you're dividing by. Search Engine Journal
2. Google Lens handles 25 billion visual searches a month, and the image is no longer the thing being optimised
Google says Lens now powers more than 25 billion visual searches every month, and one in five of them carries commercial intent. Benu Aggarwal of Milestone makes the case that AI no longer looks at a picture so much as reads it: the rooftop pool, the loungers, the lighting, the view, each one a separate clue, with the system running several searches across different parts of the image before it forms an answer. The fundamentals haven't gone anywhere (filenames, alt text, captions, transcripts, Product, Hotel, Event and ImageObject markup), but they now serve a bigger job: making sure the image, the page, the schema, the feeds and the listings are all describing the same thing.
Her word for the problem is ambiguity. When the photo, the copy and the structured data disagree, the machine has to settle the argument itself, and you no longer get a say in the verdict. Multi-location businesses are the obvious casualties: the same lovely latte photo on fourteen location pages is fourteen small ways to confuse something that's trying hard to work out which café you are. Aggarwal runs a company selling enterprise SEO software, so the advice leans towards the asset-management-at-scale end, but the principle shrinks down nicely. Pretty pictures were never really the point; now they're barely the minimum. MarTech
3. LLMs do word maths, so every rewrite of your positioning quietly moves you somewhere on the map
HubSpot has a useful plain-English explainer on vector embeddings, the lists of numbers that let AI systems match a question to a passage by meaning rather than by matching words. Ask for "the cheapest project management tool for a five-person agency" and a page about a "Starter plan built for small teams under 10 people" can still be retrieved with barely a word in common. Fractional marketer David Kirkdorffer puts it best: "LLMs do word math." There's a lovely low-tech test in there too: copy any 150-word stretch of a page into a blank document and see whether it answers anything on its own. And a timely footnote: Google stopped showing FAQ rich results and deprecated HowTo this year, so adding markup purely for its own sake buys you less than it used to.
Worth knowing HubSpot sells an AEO tool and that the headline stat (58% of marketers already optimising for answer engines) comes from its own report, so it isn't a neutral party on the urgency question. The genuinely useful bit is the warning that copy edits have consequences: swap enough words in the name of freshening things up and you can drift out of your category into the one next door. Next Net's Franklin Rios reckons retrieval is 80% the same from model to model, which argues for fundamentals over hacks, and he's politely unconvinced by the claim that you need 250 pieces of content to become citable. The thesaurus, it turns out, is now a positioning decision. HubSpot
4. TikTok and YouTube now introduce students to universities slightly more often than search engines do
The old student journey fitted neatly on a slide: see an ad, visit the website, enquire, apply, enrol. A September 2026 survey of 1,518 prospective and recently enrolled students suggests the slide needs binning. 87% had used an AI tool for education research or were considering it, and TikTok and YouTube were slightly more likely than traditional search engines to introduce them to a university they'd never previously considered. Carney doesn't say who ran the survey, so hold the exact figures a little loosely, but the shape is hard to argue with.
Choosing a university is about as high-stakes and long-winded as purchases get, exactly the sort of decision search used to own outright. If short-form video is beating Google at the ‘I'd never heard of you’ stage there, it's doing it plenty of other places too. Discovery happens wherever attention already is, and what carries you from the first glimpse to the shortlist is being recognisably the same brand on the third encounter as you were on the first. The funnel was always a diagram; now it isn't even a very good one. The Daily Carnage
5. Twelve real followers beat 120,000 bots, and the quickest way up is making other people look good
Noble Growth goes back to Kevin Kelly's 2008 essay, 1,000 True Fans, and builds five rules for organic social on top of it. First, the shortcut: bought followers don't just fail to help, they actively hurt, because 12 real people tell the algorithm who to show your post to next, and 120,000 bots tell it nothing at all. Then the rules: don't fear being cringe, be authentic, have a purpose clear enough to grasp in five seconds of scanning your profile, reciprocate (15 minutes a day making other accounts look good, and under 500 followers, embrace your inner reply guy), and be consistent. Their own case study: The National Flag Foundation, more than half a century old with no social presence at all, went from 143 Instagram followers to nearly 20,000 in four months.
It's an agency selling exactly this, and the case study is its own client, so season accordingly. Very little here is new, which is rather the point: the rules nobody writes down are mostly the ones that need discipline rather than budget. Reciprocity is the underrated one. Most accounts post into the void and wonder why nobody replies, while never once replying to anyone else. Social is still, annoyingly, social. Noble Growth
Small business shout-out
This week's shout-out goes to The Little Green Fox, a family-run eco shop in the Sussex countryside, run by Jodie around a toddler, a baby and Rufus, the self-described ‘infamous’ Weimaraner. It brings brands like Miniml, FILL, Upcircle and SESI under one roof, sends refills out in rescued plastic bottles, takes back closed-loop bottles and tubs, and packs everything without a scrap of plastic (even the tape is recyclable). It started with a few eco swaps after becoming parents in 2019, and grew into something bigger after Jodie lost her mum that Christmas: a woman who worked in a small high-street business and always told her to shop small. Consider this us passing it on. Know a small business we should feature? Reply and let us know.
