Every culture on earth has a version of this scene.
Before your grandmother did business with anyone (a butcher, a builder, a potential son-in-law) she did not read his brochure. She asked around. The neighbors, the women at the market, that one aunt who somehow knew everything about everyone. Whatever the man said about himself was politely heard and completely discounted, because your grandmother understood the oldest rule of trust there is:
Never ask the shopkeeper if his shop is good.
In Arabic it's practically a legal procedure: "اسأل عنه" — ask about him. Elsewhere it's "do your homework," "check his references," "ask around town." Different words. Same wisdom, thousands of years old: self-description is not evidence. The neighborhood is evidence.
So here is the one sentence this article exists for. AI didn't invent a new way of judging your brand. It resurrected your grandmother's way, except it asks the entire internet about you, in every language, in under three seconds, every single time a potential customer types your category into a chat box.
And most brands are still standing in front of the shop, pointing proudly at their own sign.
Your website just got demoted to testimony
For twenty-five years, the website was the brand. "Check our website" was the final word on any question. The headquarters. The source of truth. Every budget reflected it: redesign the site, perfect the site, drive traffic to the site.
That era deserves a proper burial, so let's be precise about what died. To an AI model, your website is now one witness in your own trial. And not a strong one, because it's the defendant.
Think about how these systems actually build an answer when someone asks "who's a good contractor in Dubai?" or "is this brand any good?" They don't visit your homepage and take dictation. They weigh everything at once: reviews, forum threads, news mentions, directories, comparison articles, complaint boards, your replies to angry customers, and other people's descriptions of you, which they trust considerably more than yours. Your beautiful website gets a vote. So does a Reddit thread from 2023 written by a stranger eating lunch. In the machine's arithmetic, as in your grandmother's, the person with no reason to lie tends to outweigh the brand with every reason to.
You always knew this about people. You just didn't expect the most advanced technology on earth to formalize your aunt's methodology.
The tour of things you forgot were your brand
"The whole internet" sounds abstract. It isn't. It's a very specific list of things you stopped thinking about years ago, and each of them now functions as a page of your brand book whether you like it or not.
That unanswered one-star review. Not the review itself. The unanswered part. A complaint is data; a complaint the brand ignored for eight months is a verdict. Machines read silence fluently.
Your employees' LinkedIn profiles. Your site says "a team of 50 experts." LinkedIn shows nine people, three of whom left last year. Nobody ever thought of headcount as marketing copy. It is now a fact-check, and you're failing it in public.
The forum thread you've never seen. Somewhere, someone once asked "has anyone actually used these guys?" and two strangers answered. That thread sits in your sales funnel today. You are not in it. You've never read it. The AI has.
The old address in a business directory. Trivial to you. To a machine cross-referencing facts, it's an inconsistency, and inconsistency is largely how models decide who not to recommend with confidence. Humans skim past contradictions. Machines collect them.
Your name, spelled three ways. This one is painfully regional. English on the website, one transliteration on the trade license, another in an Arabic news mention, a third in Google Maps. To you these are obviously the same company. To a model deciding whether it knows you, they might be three faint companies instead of one solid one. Your brand is being diluted by its own spelling, and nobody in the building owns that problem. Ask around your office tomorrow if you don't believe me. Watch everyone point at someone else.
The founder's five-year-old interview. You forgot it existed. The training data didn't.
Notice what that list has in common. None of it lives on your website. None of it came from your marketing department. And all of it answers the only question the machine is really asking, which is whether the neighborhood's story about you matches your story about you.
Why this was always true
Now the part that should make you smile rather than panic: nothing about human trust has actually changed.
Psychologists have known for decades that we discount self-interested claims. The moment we detect a persuasion attempt, a mental tariff gets applied to everything said afterward. It's why "award-winning" on your own homepage bounces off while a stranger's offhand "yeah, they're solid" lands like scripture. First-party claims are advertising. Third-party claims are information. Your grandmother never needed the terminology; she had the tariff pre-installed.
What changed is only the cost. "Asking around" used to be expensive. A customer had to actually call references, actually know a guy, actually have an aunt in the right neighborhood, and most people couldn't be bothered. So brands could win on self-presentation alone, and an entire industry grew up inside that loophole. Looking back, the website era was a brief, strange window in history when the shopkeeper's sign mattered more than the neighborhood, for the single reason that the neighborhood was hard to reach.
AI closed the loophole. The full background check now costs one sentence and three seconds, in any language the customer prefers. Polishing the sign while the neighborhood says nothing, or worse, stopped being a strategy. It's a costume now.
The three-second mirror
Your homework is free and takes less time than reading this did.
Open an AI assistant. Ask it about your company by name. Then by category: "best [what you do] in [your city]." Then, and this is the step your competitors will definitely skip, ask again in Arabic. Ask what people say about you. Ask who it would recommend, and why.
What comes back is what your next customer sees. Not your website. The neighborhood's testimony, compiled. For a few brands the exercise is a pleasant surprise. For most it's the audit they never commissioned: missing from the recommendations entirely, or described by a three-year-old version of themselves, or (the special regional heartbreak) present in English and a ghost in Arabic, invisible to half the neighborhood they live in.
Whatever you find, what follows is the oldest strategy there is. Stop decorating the shop. Tend the neighborhood. Answer the reviews, all of them, especially the old angry ones. Reconcile your facts everywhere your name appears, in both spellings and both languages. Show up where people actually discuss your category and be useful there, under a real name. Give the press and the directories and the forums and the models something true and consistent to say about you. They will say something either way. Your only choice is whether you contributed to it.
Your grandmother could have told you all of this, probably while insisting you eat something. Reputation was never what you say. It's what the market says back when someone asks about you.
And now someone is always asking…
Top comments (0)