Heather Holmes of Publicity For Good explains what actually builds the kind of authority that gets your business recommended by ChatGPT, Claude and Perplexity.
Every week, I talk to founders who are spending thousands on Google Ads and Meta campaigns and getting results but the moment one of their prospects opens ChatGPT and searches for a recommendation in their category, they’re nowhere. A competitor they’ve never heard of is the answer. Or nobody they know is.
That gap is not a budget problem. You can’t spend your way into an AI recommendation. ChatGPT, Claude, and Perplexity don’t surface brands because they paid for placement. They surface brands because independent, credible sources have already vouched for them. That kind of standing has to be built. And most brands haven’t started.
This piece covers how the process actually works, where most brands are falling short, and what to do about it, starting this week.
What this article covers:
- Why paid ads have no influence on organic AI recommendations
- The trust signals ChatGPT actually uses to decide which brands to mention
- Five ways to build AI visibility through earned media
- A quick audit you can run right now to see where you stand
What ChatGPT is actually doing when it recommends a brand
When someone asks ChatGPT “what’s the best project management tool for remote teams” or “who are the top PR firms for startups,” there’s no auction happening behind the scenes. No bid, no budget, no sponsored slot in the organic result. The AI is doing something much closer to research: scanning its training data and, increasingly, live web sources to find brands that have been independently verified as credible in that category.
The word that matters there is “independently.” ChatGPT discounts self-promotional content by design. Your product pages, your blog posts, your carefully written “about us” copy, the model reads all of it, but treats it the way a skeptical journalist treats a press release. It’s a starting point, not a source.
What actually moves you into a recommendation is third-party validation: a journalist who covered your company, a trade publication that reviewed your product, a podcast host who brought you on as the expert, an industry association that listed you as a resource. According to Muck Rack’s research into what AI models are actually reading, approximately 89% of the content cited by AI comes from earned media. Not ads. Not owned content. Earned.
The mistake that’s easy to miss
Most marketing teams built their playbook around search engines, where paid and organic work in parallel and spending more generally gets you more. That logic made sense when Google was where decisions started. But Google search volume is down 25% year over year and AI search on platforms like ChatGPT and Perplexity has grown over 300%. The channel is shifting and the new channel doesn’t reward ad spend the same way.
I talked recently with a coach who had done everything right on the paid side. Two years of consistent social spend, a real following, strong engagement. But when her prospects, the kind of people who research carefully before making a decision, looked her up in ChatGPT, she barely came up. Her competitors, who had been showing up in podcasts and trade press and industry features for the past couple of years, were the ones the AI knew.
Her paid strategy wasn’t wrong. She just hadn’t built the thing underneath it that AI actually looks for. And she didn’t know she needed to until the gap showed up in a sales conversation.
Five ways to build the kind of authority AI recommends
This isn’t about one big placement or a single campaign push. It’s about building a consistent layer of independent coverage that compounds over time. Here’s where to focus:
1. Get covered in publications your buyers actually read.
Feature articles, expert commentary, product reviews, founder profiles. When established publications write about you, those articles become the citations AI models trust most. One well-placed piece in an indexed trade publication carries more weight in a ChatGPT recommendation than a hundred optimized blog posts on your own site. The goal is a steady cadence of coverage, not a one-time hit. Start local and niche, then build up.
2. Show up on podcasts consistently.
Podcasts are probably the most underused tool for AI visibility right now, and they’re more accessible than most founders think. When you appear as a guest, that conversation gets transcribed, indexed, and picked up by the same crawlers that feed AI models. It creates a signal that your name and your topic consistently appear together across independent platforms, which is exactly what the AI is looking for. A client of mine who started with local and regional media appearances landed a feature on Martha Stewart’s platform four months in, and the Today Show the month after that. Each appearance built on the last.
3. Write for external platforms, not just your own blog.
Publishing a byline in a trade journal, a contributor column, or a respected industry newsletter puts your thinking in front of AI models in a format they actually weight. Your own blog has value, but it doesn’t carry the same third-party signal. LinkedIn articles are also worth doing consistently. They’re indexed by AI crawlers and contribute to the overall picture of your expertise. My book, Seen by AI, Found by Customers, goes deeper on how to structure thought leadership content for AI visibility specifically.
4. Make your social proof specific and findable.
AI models do pull testimonials, case studies, and third-party endorsements but only when they’re structured clearly and attributed to real outcomes. Vague quotes don’t register the same way specific results do. A client who grew her business by 40% after a media campaign. A brand that saw 3,000% more sales on a single awareness day. A company that captured 84% of their entire category’s media mentions in sixteen months. Those numbers are what the AI finds and cites. Keep your case studies current, specific, and easy to read.
5. Clean up your brand data everywhere it lives.
This is the step most brands skip, and it matters more than people realize. AI verifies brand information by cross-referencing directories, Google Business Profile, association listings, and social bios. If your company name, description, or contact information is inconsistent across those sources, the model’s confidence in your brand drops and so does your visibility. A full audit of this usually takes less than an hour and can make a real difference in how AI tools represent you.
A quick audit to run right now
Before anything else, you need to know what the AI actually thinks of you. Open ChatGPT, Claude, and Perplexity and search three things: your brand name, your category, and the specific problem you solve. Write down what comes back.
If you show up, check whether the description is accurate. If you don’t, look at who does and spend fifteen minutes looking at their media history. You’ll almost always find the same thing: consistent, long-term earned coverage across multiple independent sources.
That’s your roadmap. Not a reason to panic, but a clear picture of what needs to be built.
This compounds in a way ads don’t
The thing about earned media that I’ve watched play out with hundreds of clients is that it doesn’t stop working when you stop paying for it. A feature article published two years ago is still being crawled. A podcast episode from eighteen months back is still indexed. Every piece of coverage adds to a body of independent evidence that the AI can draw from whenever someone asks about your category.
I’ve seen a single well-timed local TV segment bring in $112,000 in sales for a small brand with a strong story. I’ve watched a purpose-driven beverage company go from three-state distribution to a national exit in sixteen months because the media coverage compounded into something the whole industry noticed.
Ads are a dial you can turn up and down. Earned authority is infrastructure. And right now, infrastructure is what the AI is reading.
Keep up to date with our stories on LinkedIn, Twitter, Facebook and Instagram.
