AI recommends businesses on review volume, not star rating
There's a slightly deflating thread running through everything that crossed my desk this week, and it's bad news for anyone still chasing the shiny numbers. AI recommends local businesses on review volume, not on their star rating. ChatGPT cites Reddit and Wikipedia four times more often than it cites the press. The social accounts that actually take off are the ones brave enough to be a bit cringe, not a bit polished. And the pages quietly doing the selling aren't your clever blog posts, they're the boring "X vs Y" ones. Oh, and ChatGPT now has a proper ads manager, with OpenAI dropping the $50k cover charge so small businesses can finally get through the door. Here are your five.
1. AI recommends businesses on review volume, not star rating (which stings a bit if you've been chasing five stars)
We ran 10,000 US local businesses through ChatGPT and Perplexity to see what actually makes AI put a business forward, and the answer is almost rudely simple: sheer review volume. The businesses AI recommended averaged 133.4 Google reviews; the ones it ignored averaged 10.7, a twelvefold gap. Star ratings, the thing everyone agonises over, barely moved (3.8 versus 3.3 across the tiers). AI, it turns out, cares far more that a lot of people bothered to review you than that every single one left a perfect score.
So the shop with 150 reviews at a slightly wobbly 4.2 quietly out-ranks the one guarding its pristine 4.9 across twenty. Reviews, not vibes. If you've been filing review-gathering under reputation admin rather than the single biggest lever on whether AI mentions you at all, this is your nudge to swap the two round. Insites
2. ChatGPT now has an ads manager, and OpenAI just scrapped the $50k door charge for small businesses
ChatGPT ads have gone self-serve across the US, and the interesting part isn't the ads, it's who's suddenly allowed to buy them. OpenAI has binned the $50,000 minimum spend from its pilot, the sort of number that politely tells a small business to come back never. Now everyone walks through the same door: SMBs, startups and the big holding companies (Dentsu, Omnicom, Publicis, WPP) alike. Cost-per-click bidding and a conversion pixel are live today, while a Conversions API, CPA bidding and third-party measurement are all filed under "coming", though OpenAI's own monetisation lead admits there's no partner and no timeline for that last one yet.
For now the categories are fenced off to safe-ish territory (household and consumer goods, local services, travel, entertainment, digital products and education), and OpenAI keeps every placement decision in-house rather than handing the keys to ad tech. Translation: the channel is real and finally open to normal budgets, but the measurement you'd want before pouring money in is still a promise rather than a feature. Early days, then, but worth watching now the door's unlocked. Digiday
3. Your best-converting pages aren't blog posts, they're the boring "X vs Y" and "best alternatives" ones
Here's a stat to reframe the content calendar: the top 20 CRM brands alone pull around 178,000 searches a month between them, and nearly all of it is bottom-of-funnel, people typing "X alternatives" or "Tool A vs Tool B" with a card already half out of their pocket. MRR Unlocked lays out the three page types that catch this demand (alternatives lists, category "best of" guides, and head-to-head comparisons) and does the napkin maths: 20,000 high-intent visitors a month at a 3% conversion is 600-odd leads, from pages most brands never get round to writing. One example comparison page hit #4 in the US inside 14 days.
The catch nobody mentions is that these pages get written with or without you: skip the comparison and a competitor writes it, or an AI confidently guesses, and neither version will be flattering. Bottom-of-funnel content has a reputation for being dull, which is exactly why it's underbuilt and quietly overperforming. The unglamorous "vs" page is doing more selling than the thought-leadership piece you agonised over. MRR Unlocked
4. When ChatGPT cites a source, it's four times more likely to be Reddit or Wikipedia than the press
Someone combed through 35,000 ChatGPT citations for SaaS queries and found that once you strip out vendors' own sites, user-generated content (Reddit threads, Wikipedia pages, LinkedIn posts) is the biggest third-party source by a mile: 17.1% of cited domains, more than four times the 4.0% that traditional publishers manage. Wikipedia does most of the heavy lifting, Reddit swings about wildly (down 11.7% one month, up 18.1% the next), and LinkedIn quietly grew its citation share 43.3% in late June. The community, in other words, is writing AI's reference list.
Two things stand out. Named humans get cited more than faceless brand accounts, which is a decent argument for putting real people's names on things. And Google's shiny new social metrics in Search Console conveniently leave out Reddit and LinkedIn, the two sources doing the most work, so the official numbers will always understate where AI is really forming its opinions. Turns out the strangers on the internet you were warned about are the ones the robots trust most. Search Engine Land
5. One account went from 143 to 20,000 followers in four months, and rule one was "don't fear being cringe"
Noble Growth rounded up the unwritten rules of organic social and led with the least corporate advice imaginable: don't fear being cringe. Post often, post confidently, and accept that plenty won't land, because the account publishing daily simply gets more shots on goal than the one polishing a monthly masterpiece. The proof sits in a charity, the National Flag Foundation, that went from 143 Instagram followers to nearly 20,000 in four months doing exactly that. The rest of the rules are almost aggressively unsexy: be authentic, pick a niche you can explain in one sentence, actually engage with other accounts, and show up every single day.
There's a Kevin Kelly essay from 2008 buried under all this, "1,000 True Fans", and the whole thing rhymes neatly with the review and AI-citation stories above: real, engaged humans beat big vanity numbers every time. It's also why buying followers backfires, since bots don't behave like an audience and the algorithm notices the difference and quietly throttles your reach. A thousand people who actually care will out-earn a hundred thousand who don't. Noble Growth
Small business shout-out
This week's shout-out goes to Peace of Heritage, which makes felted flower hair clips by hand, each one modelled on a real bloom from Caribbean and West African heritage: sorrel, sea hibiscus, blue mahoe, the pride of Barbados. No mass-produced plastic, just wool, patience and a lot of colour, with bespoke pieces made to order around whichever flower carries meaning for you.
Based in the UK and shipping to more than 30 countries (Barbados, Jamaica and Grenada included), the brand's whole thing is "crafting stories of heritage and the land", turning a hair accessory into a small, wearable piece of where you're from. This is technically a micro-business, so we'd love to know if you purchase from them! Know a small business we should feature? Reply and let us know.
