If you’ve spent any time thinking about GEO, you’ve probably noticed that earned media keeps coming up. There’s a good reason for that. According to research from Observer, 89% of links cited by AI originate from earned media sources. Your own website content matters, but AI engines overwhelmingly lean on what others have written about you.
The natural follow-up question is: how much earned media do you actually need? A couple of mentions? A dozen? A sustained PR campaign over months?
The honest answer is that it depends entirely on your category. Some niches are so editorially saturated that breaking through requires serious, sustained investment. Others are wide open, with AI engines still figuring out who to recommend.
We call this dynamic the editorial tax – a framework we’ve developed at PolyGrowth to help clients understand what they’re up against before they invest in GEO.
What is the editorial tax?
The editorial tax is the threshold of third-party editorial coverage a brand must accumulate before AI engines treat it as a credible recommendation. Think of it like a minimum buy-in at a poker table – you can’t play until you’ve met it, and the amount varies by table.
Continue reading: 5 Steps to Build External Validation for AI Recommendations
This threshold exists because of how large language models form their outputs. AI engines don’t rank pages the way Google does. They synthesise answers from patterns across their training data and retrieval sources.
When multiple authoritative publications mention a brand in similar contexts, the model develops what you might call statistical confidence – enough signal to surface that brand as a recommendation. Below that confidence level, a brand simply doesn’t appear, regardless of how good its product is.
An Ahrefs study of AI search overlap found that only 12% of AI-cited URLs rank in Google’s top 10, and 80% don’t rank anywhere in Google’s top 100. AI engines are drawing from a fundamentally different corpus than traditional search, one that’s heavily weighted toward editorial and media sources rather than SEO-optimised pages.
The correlation between editorial presence and AI visibility is striking. A separate Ahrefs analysis of 75,000 brands showed that brands in the top 25% for web mentions earn over 10× more AI Overview mentions than the next quartile, with web mentions correlating 0.664 with AI visibility.
Research from Surfer SEO across 289,000 URLs reinforces this: AI is significantly more likely to strongly recommend a brand when it’s mentioned across many cited source pages, with a Spearman correlation of 0.41.
What makes the editorial tax particularly tricky is that it differs across AI engines. ChatGPT, Gemini, and Claude each weight sources differently and have distinct consensus behaviours. A brand might clear the threshold on one engine but remain invisible on another, depending on which publications each model prioritises in its retrieval and training pipeline.
What our research shows about editorial tax across categories
We’ve studied editorial tax dynamics across both SaaS categories and B2B service industries, alongside other fields. The variation is enormous, and the pattern is consistent: editorial density determines AI consensus.
In our SaaS study, email marketing stood out as the most editorially dense category we tested. Hundreds of “best email marketing tools” articles exist across major publications, and the same five to seven brands appeared consistently across ChatGPT, Gemini, and Claude.
Continue reading: How to Get Your SaaS Startup Recommended by AI
The AI engines had so much editorial signal to work with that they’d essentially reached consensus. For a newcomer trying to break into that list, the editorial tax is steep. You’d need sustained, high-quality coverage across multiple authoritative publications just to register.
Contrast that with web design in our B2B services study: The editorial ecosystem is thin – far fewer roundup articles, fewer “best of” lists, and limited media attention compared to SaaS.
The result? Near-zero agreement between engines. We measured Jaccard similarity scores of just 0.05–0.10 across ChatGPT, Gemini, and Claude, meaning the engines were recommending almost entirely different brands. We scored web design as a 5/5 GEO opportunity – wide open for any firm willing to invest.
Recruitment tells the opposite story. Hays accumulated roughly 29 AI mentions across engines, Reed had 17, and Michael Page had 16. These brands showed up consistently across all three engines, locking them into the top spots. We scored recruitment as a 1/5 GEO opportunity – the editorial tax has already been paid by the incumbents, and displacing them would require substantial and sustained media investment.
The takeaway is straightforward: the more editorially dense your category, the higher the editorial tax for newcomers, and the harder it is to shift AI consensus once it’s formed.
The threshold in action – a client case study
We saw the editorial tax play out in real time with a client in a competitive SaaS and commerce niche. This was an established brand – strong organic search rankings, excellent Google ratings, a well-known name in their space. By any traditional metric, they were doing well.
But across ChatGPT, Gemini, and Claude, they were invisible. Zero AI visibility despite being a credible, well-regarded player.
The diagnosis was clear: they had virtually no PR presence and no meaningful third-party editorial coverage. Their brand existed in their own ecosystem – their website, their blog, their customers’ word of mouth – but not in the publications that AI engines rely on.
We began a sustained earned media programme. Over several months, we secured top tier thought leadership quotes, placements in listicle articles, and reviews on reputable industry publications. The impact on AI visibility was measurable yet marginal. Each placement added signal, but the cumulative weight hadn’t yet crossed the threshold.
Then, in early 2026, we landed a major feature in a top-tier national media outlet. Within weeks, the shift was visible across the board. All three AI engines and Google AI Overviews began recommending the client with noticeably higher frequency. In quite a few query contexts, the engines even named them as the leading option in their niche.
What happened next was even more revealing. AI engines started citing the client’s own blog content and website pages – material that had existed for months but had never been surfaced in AI responses.
The top-tier media feature appeared to act as a credibility signal that unlocked the entire content footprint. It was as though the models suddenly “trusted” the brand enough to draw from its own resources.
This is the editorial tax in action. The expert quotes and listicle mentions were necessary groundwork, but they weren’t sufficient on their own. It took a high-authority feature to push the brand past the threshold. Once crossed, the compound effect kicked in – not just from that single outlet, but across every engine simultaneously.
The editorial tax is ongoing
Crossing the threshold once doesn’t guarantee you stay there. AI training data evolves, models are retrained on fresh corpora, and competitors are earning their own coverage in the meantime.
We continue to invest in earned media for the client above, treating it as a subscription rather than a campaign with a defined end date. The editorial tax isn’t a one-off fee – it’s closer to a recurring cost of maintaining AI visibility.
The brands dominating AI recommendations in mature categories like email marketing or recruitment haven’t simply earned their position. They’ve maintained it through years of accumulated editorial presence – conference coverage, analyst mentions, niche media and blog features, and relationships that each add a piece to their media puzzle.
What this means for your GEO strategy
Before investing in GEO, understand your category’s editorial tax. The GEO opportunity scoring we use in our research – from 1/5 (locked) to 5/5 (wide open) – is a practical way to gauge whether breaking through is feasible and what level of media investment it’s likely to require.
Our client case study showed that one top-tier placement can do more than a dozen mid-tier ones. Still, that doesn’t mean that the smaller features we had secured earlier didn’t add to a bigger picture. Most likely, this is about a combination of volume and quality (although we should clearly state that volume doesn’t mean anything goes).
Furthermore, treat earned media as an ongoing channel, not something you switch on for a quarter and then abandon. AI visibility is a moving target, and the brands that hold their positions are the ones that keep feeding the editorial ecosystem around them.
Last but not least, if the editorial tax in your field is exceptionally high, consider going for more specific niches. Look into specific geographies, features, target customers, or other aspects – you will find that the editorial tax is significantly lower for these sub-niches.
Frequently asked questions
How do I know what my category’s editorial tax is?
Start by querying ChatGPT, Gemini, and Claude with the same prompts your potential customers would use – things like “best [category] tools” or “top [service] providers.” Compare the results across engines.
If the same brands appear consistently, you’re looking at a high editorial tax category with strong AI consensus. If the results are fragmented and different on each engine, the tax is lower and the opportunity is larger.
Continue reading: 7 Steps to Get a New Brand Into AI Search Results From Scratch
Our GEO opportunity scoring framework formalises this analysis, but even a manual comparison gives you a useful starting point.
Does the editorial tax apply to all industries and niches?
The concept applies universally, but the practical level varies enormously. B2B SaaS categories tend to have well-developed editorial ecosystems with extensive “best of” coverage, which means higher editorial taxes.
Niche professional services or emerging categories often have thinner coverage, making them easier to break into. For some micro niches, such as businesses in very small countries, you may find that no media coverage is needed at all to be recommended by AI tools.
The key variable is how much third-party editorial content exists in your space – the more that’s out there, the more you need to compete with.
How important is earned media vs. my own website’s content?
Both matter, but they serve different roles. Our client case study showed that owned content – blog posts, landing pages, product documentation – can be surfaced by AI engines, but typically only after enough earned media coverage establishes the brand’s credibility.
Think of earned media as the unlock and owned content as the inventory. Without the editorial credibility signal, AI engines tend to overlook your own content no matter how strong it is.
Continue reading: How ChatGPT Decides Which Businesses to Recommend
Do directory listings like Clutch or G2 count toward the editorial tax?
They contribute, but they carry less weight than traditional earned media. AI engines do reference review platforms, and strong presence on Clutch or G2 can add signal – particularly when those listings include detailed reviews and rankings.
However, our research suggests that editorial coverage in recognised publications, media outlets, and industry-specific content sites carries more influence on AI recommendations than directory listings alone. Directory presence is a useful complement, not a substitute for earned media.
