AI and the Impact of LLMs on the Digital Publisher Ecosystem

At CES this year, no theme stood out more than AI and its impact. It’s awfully easy to get caught up in the hype of all things AI, but I try to take a pragmatic approach: look for underlying KPIs and pressure-test those against future scenarios. One of the KPIs for the digital ecosystem is the impact of AI on digital publisher health.

The fundamental issue is that LLM searches, at least as of today, do not drive web traffic to publishers like the legacy Google model did via algorithmic and paid ranking of publisher links.  Today there is no direct equivalent of Google links in LLM search results, as LLMs endeavor to answer the user question/query with an actual answer vs links to content (publishers) that provide that answer. This is the root of what we all experience as a massively superior user experience, which has driven the rapid adoption of LLMs.   

Just how fast have LLMs grown?

All of this is to say that LLMs have grown quickly. Historically quickly.

The next question is the impact on web publishers so far. This was the major topic in our industry gatherings at CES (and for the last year or so).

The impact of LLMs on publisher traffic

You have to do some estimating around market share of LLMs vs search. Strangely, the measurement firms don’t consider LLM searches to be in the same category of web search queries, but alas, there’s ChatGPT to the rescue with this analysis below.

If you normalize web search and LLMs search to one KPI, LLMs capture between 15% and 17% of total search volume today. But there is not a 1:1 loss correlation with publishers, since some LLM searches do ultimately provide links and referrals to web publishers. This is estimated at between 1% and 2%. Other analysis suggests that Google has reduced to 78% and ChatGPT at 17%, with other LLMs making up the difference.

If you then project out the growth of LLMs and the impact of that growth on publisher referrals, here’s what you get:

And yet some research has suggested that web traffic referrals are materially worse (eg., 40%+), and anecdotal feedback from publishers and SSPs from CES discussions suggest the reduction could be as high as 30%+ YoY already.

So what does the future look like if we assume LLMs will continue to capture consumer market share over web search?

So in summary, LLMs are decreasing search referrals to publishers today at no less than 15%. And that number is set to increase to ~50% in the next 4 years. I’m no math major, but as a pragmatist, that’s not so good for the open internet economy. Without another solution, performance advertisers would grow even more dependent on the walled gardens, emboldening the walled gardens to raise prices and ignore calls for more transparency. If performance marketers have nowhere else to go, the walled gardens will benefit at the expense of all the open internet publishers and advertisers.

How publishers can respond

What will publishers do to offset this alarming threat to their existence?

Let’s start with confronting reality. There is a clear axiom in the media business: a direct relationship with the audience is the most strategic asset a media company (publisher) can have.

And yet, publishers have become dependent on indirect channels to drive audience growth via search referrals. They have also become lazy in terms of building non-search distribution channels. Unlike Performance/D2C marketers, traditional web publishers have not built core competence around identifying and building growth channels outside of search.

To be blunt, publishers need to step out of the search referral guillotine line and go to work building direct relationships with consumers. They need to become growth marketers. 

Publishers should shift from search dependency to a diversified strategy combining owned audiences, multi-platform distribution, AI-friendly content optimization, brand growth marketing, and diversified revenue models to maintain visibility and revenue in an AI-driven discovery world.

1. Own direct audience channels (high impact):

Grow email newsletters, mobile apps, and push/SMS subscriptions to build ownable traffic. These channels convert at higher rates and are less vulnerable to AI search shifts; many publishers with strong newsletters still grow revenue despite traffic declines.

2. Embrace LLMs (because they are here for good)

1. Put a fence around your owned content to prevent LLMs from using it without permission. Cloudflare is a leading tech provider in this category.
2. License content to LLMs. This will ultimately be a marketplace-driven process where LLMs pay content creators/owners a share of downstream monetization based on the usage of licensed content (ProRata.ai does this today with dozens of tier-one publishers). You can also cut shorter-term deals with LLMs directly (since there are no standards yet and you don’t want to lock in a long-term bad deal).
3. Build your own LLM to provide verticalized LLM search for your consumers, and integrate it within your content. ProRata.ai has a specific solution, or you can build on top of the leading LLMs.

3. Optimize for AI/LLM discovery (medium to high impact):
Use Generative Engine Optimization (GEO)/Answer Engine Optimization (AEO) techniques so AI systems cite your content rather than just summarizing it. This may not restore all lost clicks, but it increases brand presence inside emerging discovery paradigms.

4. Diversify revenue streams (high impact):
Shift away from ad revenue tied to referral traffic:

• Subscriptions and memberships
• Commerce/affiliate integrations
• Events and premium digital products

Publishers with strong subscription offers often sustain or grow revenue even as organic traffic falls.

5. Brand & growth marketing (medium to high impact):
Invest in proactive brand building through CTV and mobile app acquisition campaigns and social/influencer partnerships. Brand strength reduces dependence on SEO for discovery and drives direct visitation. And invest in building hard core growth marketing expertise in the model of D2C or mobile app growth marketing companies. 

6. Leverage first-party data & personalization (medium impact):
Use analytics from direct channels to tailor experiences and recommendations, increasing engagement and monetization per visitor. First-party data also underpins more effective advertising offers and audience segmentation.

Where TV sits — and what to learn from it

As a CEO and Founder of a Performance TV company, I must ask: where does TV play into this rapidly evolving world, and can TV help publishers mitigate the LLM threat in any way?

The answer is perhaps best illustrated by a question: when was the last time we saw a major publisher advertise on TV? I personally have never seen a NYT TV ad, for example. I have not seen a TV ad for most of the global top 20 publishers.  

This tells us that publishers don’t see investing in TV (top of funnel or bottom of funnel) as a valuable marketing channel. For businesses that rely on consumer relationships, this fact is striking.

I see the issue completely differently. TV can drive value in a world where publishers are in decline due to LLMs in several key areas:

1. TV can help build a direct relationship with a publisher’s target audience.

From brand building to performance advertising, where a subscription might be the outcome KPI, TV advertising can be a huge lever to help publishers grow.

2. For advertisers, a diminished open internet will only add additional reliance on walled gardens. 

And as we have discussed for years, marketers that build reliance on black box walled gardens are setting themselves up for the dynamic that can best be described by the infamous Bezos mantra: “Your margin is our opportunity.” Marketers can’t build a business without basic transparency, real KPIs, or concrete customer IDs.

TV can offset this potential dark future by helping advertisers (and publishers) build direct relationships with consumers in a fully transparent manner. In turn, this leads to fundamental marketing capabilities around audience growth and retention, which in turn drives enterprise value.