Something happened last Tuesday that made me rethink everything I thought I knew about word-of-mouth marketing.
A friend of mine -- she runs a mid-size accounting firm in Toronto -- told me she got four new clients in one month, all of them saying the same thing: "ChatGPT recommended you." Not Google. Not a referral from a colleague. An AI chatbot, apparently, decided that her firm was worth mentioning when someone asked "best accounting firms for small businesses in Toronto."
She didn't do anything special for this. She had no idea it was happening. And that, right there, is the problem and the opportunity all at once.
The new word-of-mouth is artificial
Let's be honest about what is happening here. Millions of people -- and the number grows every month -- have started treating ChatGPT and Perplexity the way they used to treat Google. They ask questions, they get answers, and those answers include specific business names.
A Gartner study projected that by 2026, traditional search engine volume would drop by 25% as users shifted to AI-powered answers. We're living in that projection now. When someone types "best Italian restaurant near the financial district" into Perplexity, they don't get ten blue links. They get a direct answer: "Based on reviews and local reputation, these are the top options..." followed by three or four names.
Your business is either one of those names, or it is not. There is no page two.
The truth is, this represents a fundamental change in how recommendations work. For decades, word-of-mouth was human. Your cousin told you about a good plumber. Your colleague mentioned a reliable CRM. Now an AI system trained on the entire internet is doing the recommending, and it reaches more people in a day than your cousin reaches in a lifetime.
How AI models decide who to mention
I think most business owners have a completely wrong mental model of how this works. They imagine there's some list somewhere, some directory they can submit to, some ad they can buy. There isn't.
ChatGPT and Perplexity build their understanding of the world from text. Enormous quantities of text. Training data includes websites, articles, forums, reviews, news stories, social media discussions -- basically everything that's been written on the internet. When someone asks "who makes good project management software," the AI doesn't look up a database. It reconstructs an answer based on patterns it absorbed during training.
This means the AI's recommendation is shaped by how often and in what context your brand appears across the web. If your company name shows up repeatedly in positive contexts related to your industry, the model learns to associate your brand with that category. If you barely exist online, or only exist on your own website, the model may not know you exist at all.
Perplexity works slightly differently -- it searches the live web and synthesizes results in real time. So the signals that matter for Perplexity overlap with those that matter for ChatGPT, but with more weight on current, crawlable web presence.
There's a study from late 2025 that looked at 75,000 brands and found brand mentions had roughly a 0.65 correlation with appearance in AI-generated answers. Backlinks, by contrast, showed weak or neutral correlation. The implication is hard to ignore: being talked about matters more than being linked to.
The seven things that make AI recommend you
I've been tracking this for months now, and there are clear patterns in which businesses AI systems tend to mention. Not all of these carry equal weight, and their importance shifts depending on your industry. But collectively, they paint a picture of what "AI-visible" looks like.
1. A strong, consistent brand entity across the web
This one sounds obvious but most businesses get it wrong. Your brand needs to exist as a recognizable entity -- the same name, the same description, the same core identity -- across many different places online.
Think about it from the AI's perspective. If you're called "Smith & Associates" on your website, "Smith Associates LLC" on LinkedIn, "Smith Assoc" in a directory listing, and "John Smith's Tax Practice" on Yelp, the AI may not connect all of these as the same business. You've fragmented your own brand signal.
The businesses that get mentioned consistently have consistent naming. They have a Wikipedia page, or at least a Wikidata entry. They have a Google Business Profile that matches their website. Their About page tells a clear story about what they do and for whom. Every mention across the web reinforces the same identity.
A practical test: Google your business name in quotes. Do all the results clearly refer to the same entity? Or is there ambiguity?
2. Being cited by authoritative sources
AI models don't treat all mentions equally. A mention in Forbes or a well-known industry publication carries far more weight than a mention on a random blog with twelve readers.
This is where PR stops being a "nice to have" and becomes a strategic necessity. Getting your CEO quoted in an industry article, having your company included in a research report, being mentioned in a news story -- all of this builds the kind of authoritative signal that AI systems learn from.
I spoke with a marketing director at a SaaS company last month who told me they'd shifted 30% of their link-building budget toward earned media. Not for the links anymore -- for the mentions. "We stopped caring about whether the journalist included a hyperlink," she said. "We care about whether they named us." That's the right instinct.
3. Active presence on review platforms
Reviews are training data. Full stop.
When ChatGPT learns about businesses, it learns from Yelp reviews, Google reviews, G2 reviews, Trustpilot, industry-specific review platforms -- all of it. A business with 400 Google reviews and a 4.6 average rating occupies more "mental space" in the model than a business with 12 reviews and a 4.8.
Volume matters, not just score. And recency matters too. A burst of great reviews from 2023 followed by silence is less compelling than a steady stream of reviews coming in regularly.
One thing I've noticed in testing: Perplexity in particular seems to weight review platforms heavily when answering local business queries. If you ask it for recommendations, it often explicitly references review scores and counts. So your review strategy isn't just about Google rankings anymore -- it's about AI recommendations.
4. Structured data and schema markup
This is the technical piece, and it matters more than most people realize.
Schema markup -- that structured data you add to your website's code -- helps AI systems understand exactly what your business is, what you offer, where you're located, and how to categorize you. It's like filling out a form that AI systems can read.
LocalBusiness schema, Organization schema, Product schema, FAQ schema, Review schema -- these aren't just for Google's rich snippets anymore. They create machine-readable descriptions of your business that AI crawlers can extract and use.
I tested this with a client site last year. We added comprehensive schema markup -- Organization, LocalBusiness, all their services as Service schema, FAQ schema on every major page. Within two months, their mentions in Perplexity answers for industry queries went from zero to appearing in roughly one out of four relevant queries we tested. Correlation isn't causation, naturally, but the timing was hard to ignore.
5. Content that answers specific questions
This is where content strategy meets AI visibility, and where most businesses get it fundamentally wrong.
Most business blogs write about themselves. "We're proud to announce our new feature." "Our team attended the industry conference." Nobody is asking ChatGPT about your press releases.
What people are asking: "What's the best way to handle payroll for remote employees?" or "How do I choose between cloud and on-premise CRM?" or "What should I look for in a commercial lease?"
If your website has authoritative, well-written content that answers these specific questions -- and your brand is naturally mentioned in that content as the entity providing this expertise -- AI systems learn to associate you with those topics.
The format matters too. Clear question-and-answer formats, well-structured headings, concise authoritative paragraphs -- these are all easier for AI systems to extract and quote. Perplexity in particular loves content it can directly cite and attribute.
6. Mentions in forums, Reddit, and Quora
Here's something that surprises most business owners: Reddit threads are training data. Quora answers are training data. Forum discussions are training data.
When someone asks on Reddit "what CRM do you guys use for a 20-person sales team?" and three people reply with your company name and a positive comment, that's a powerful signal. It's organic, it's in context, it's associated with a specific use case, and it appears in a format that AI models are specifically trained on.
I'm not suggesting you go astroturf Reddit. That backfires spectacularly and people can smell it. But I am suggesting you pay attention to where your brand appears in community discussions. Are your customers talking about you? Are they recommending you when someone asks for advice?
If the answer is no, you have a community engagement problem, not just an AI visibility problem. The businesses that get mentioned by ChatGPT in niche queries are usually the ones that actual humans were already recommending in forums and discussion threads.
Some companies encourage this organically by building communities around their product, by having team members who genuinely participate in industry discussions (as themselves, not as corporate mouthpieces), and by making their product good enough that people voluntarily recommend it. There is no shortcut for that last part.
7. Fresh, regularly updated content
AI models get retrained and updated. Perplexity crawls the live web. Both systems have a recency bias -- more recent information is generally weighted more heavily than older content.
A website that hasn't published anything new since 2024 is at a disadvantage compared to one that publishes weekly. Not because publishing frequently is inherently virtuous, but because fresh content signals that the business is active, current, and relevant.
This applies to your core website pages too, not just blog posts. When was the last time you updated your service descriptions? Your About page? Your FAQ section? Stale content sends a signal -- to both humans and AI systems -- that nobody's paying attention.
I keep seeing businesses with beautiful websites built three years ago that haven't changed a word since launch. Those sites are fading from AI awareness while competitors who update regularly are becoming more visible.
What not to do
Let me be direct about some approaches that waste time or actively hurt you.
Trying to manipulate AI prompts is pointless. I've seen services offering to "optimize your business for ChatGPT" by somehow gaming the prompt system. This is nonsense. You don't control what users type, and even if you could, ChatGPT's responses are based on its training data, not on some magic phrase.
Fake reviews are destructive. AI systems are getting better at detecting patterns associated with inauthentic reviews. And Perplexity, which cites sources in real time, can surface the platforms where your reviews live. If those reviews look manufactured, that's what gets associated with your brand.
Keyword stuffing your website doesn't work either. Writing "best plumber in Chicago" seventeen times on your homepage didn't work for Google in 2015, and it doesn't work for AI in 2026. AI systems understand language contextually. They can tell the difference between content written to inform and content written to manipulate.
Creating dozens of thin pages targeting every possible query is another trap. Some SEO agencies are now pitching "AI optimization" packages that involve creating hundreds of pages with slightly different variations of questions. AI systems are trained to synthesize information, not count pages. One genuinely authoritative page beats fifty thin ones.
How to test where you stand right now
This part is easy and I think everyone reading this should do it today.
Open ChatGPT. Open Perplexity. Ask questions that your ideal customer would ask. Not questions about your brand -- questions about your category.
If you're a law firm: "What are the best employment lawyers in [your city]?" If you sell software: "What tools are best for [the problem you solve]?" If you're a restaurant: "Where should I eat [your cuisine] in [your area]?"
Do you appear? Do your competitors appear? What does the AI say about you versus about them?
Try different phrasings. Try adding qualifiers like "for small businesses" or "with good reviews" or "affordable" or "premium." See if different framings bring up different names.
Write down what you find. This is your baseline. The names that appear in these answers are, right now, capturing business that might otherwise come to you. And this will only become more significant over time as AI usage grows.
The monitoring problem
Here's what frustrates me about this whole situation. With Google, you have Search Console. You know exactly which queries brought people to your site, how often you appeared, what your click-through rate was. You can track this weekly, monthly, quarterly. You have data.
With AI mentions? You're mostly flying blind.
There's no "ChatGPT Console" that tells you how often your brand was mentioned in AI responses. You don't know when someone asks Perplexity about your industry and your competitor gets recommended instead of you. The visibility gap is enormous and it's one of the biggest challenges in this space right now.
This is exactly the kind of problem we're working on at Licheo with our Generative Engine Optimization tools. Tracking how your brand appears in AI-generated answers, understanding which queries trigger mentions of your competitors, and measuring your AI visibility over time. The monitoring piece is, I believe, going to become just as important as traditional rank tracking was for SEO.
Because you cannot improve what you cannot measure. And right now, most businesses aren't measuring this at all.
What to do this week
I want to end with something practical. Not a grand strategy, not a six-month roadmap -- just what you can do in the next seven days to start building AI visibility.
First, do the audit I described above. Spend thirty minutes asking ChatGPT and Perplexity about your industry, your services, your local area. Write down every business name that appears. Note where you show up and where you don't. This takes half an hour and gives you a clear picture of reality.
Second, check your brand consistency. Google your business name in quotes and look at the first twenty results. Is your name consistent everywhere? Fix any mismatches on profiles, directories, and listings you control. This is tedious work but it matters.
Third, look at your review situation. How many Google reviews do you have? When was the last one? If it's been more than a month, you have a gap. Set up a simple process -- even just an email to happy customers asking them to leave a review. Consistency beats volume in the long run.
Fourth, update your schema markup. If you don't have Organization and LocalBusiness schema on your website, add it. If you do have it, check that it's complete and accurate. Google's Structured Data Testing Tool can validate what you have. This is a one-time technical task that pays dividends.
Fifth, identify one authoritative publication in your industry and figure out how to get mentioned there. Maybe it's a contributed article. Maybe it's responding to a journalist's query through HARO or Qwoted. Maybe it's sponsoring or speaking at an industry event that gets covered. One authoritative mention can shift your AI visibility more than fifty low-quality ones.
Sixth, look at your content. Does your website answer the questions your customers actually ask? Not marketing fluff -- real, specific, useful answers to real questions. If it doesn't, pick the five most common questions your sales team hears and write genuine, authoritative answers to them. Put them on your site.
That's six things. Not three, not five, not ten. Six things that, done well, will start building the signal that makes AI systems recognize and recommend your business. None of them require expensive tools. None of them are tricks. All of them are things that also happen to make your business more visible to humans too.
Which, alla fine, is the whole point. The businesses that AI recommends are generally the businesses that have built genuine authority and reputation online. There's no hack for that. There's only the work of actually being good at what you do and making sure the internet knows about it.