The Influenceable Web

The Influenceable Web started with one question: is traditional search shrinking?

The answer needed more than search data. It also needed to cover whether people were moving to social and niche communities, how big the AI-driven search and conversational tools had become, where else attention was going, and how a marketer could measure any of it.

So I built the dataset from scratch. I hand-pulled clickstream data from Similarweb's new DataHub beta, after I ran out of Data credits. The result is ~100 deliberately chosen websites that show how many people visit browser-based sites each month, going back to March 2022, and which parts of the web a marketer can still influence. I maintain it by hand, every month.

Why I care

If the zero-click future is real, and the data says it is, the work moves from owned surfaces to earned and discovery surfaces. I wanted to map the places on the web where three conditions hold:

• A marketer can do marketing there.
• The audience is receptive to being influenced.
• It's measurable: the audience is browser-based.

That's the Influenceable Web.

What the data says

The Influenceable Web contracted about 3% in 2025 versus 2024. The audiences inside it are moving into in-app surfaces, what older analytics decks would have called "dark." A buyer's path used to run through surfaces you could track: search, your site, conversion. Today the same buyer might start in iMessage, move through the Reddit app, validate the answer in ChatGPT, and transact in a marketplace app. None of those surfaces show up in a standard analytics stack.

The chart below maps every major category of the Influenceable Web: the browser-based surfaces where marketers can reach receptive audiences. It tracks unique monthly visitors across nine primary categories from April 2022 through April 2025.

How I read the data

If a metric needs three caveats to explain itself, shelve it and find one your CFO will sit through. Blend the sources. I'm buying a bundle of clickstream, platform data, and server logs, because no single source is honest enough on its own. Track relative movements. Month over month is where the signal lives. And read the category-level shifts. If Video and Streaming is climbing inside your audience while Reference and Forums is flat, that's the trend worth acting on.

What I'm doing next

Auditing how my content renders, because most LLMs ignore client-rendered pages. Logging referrers at the edge so I can catch the user agents that show up. Building entity-dense content to land in AI Overviews, AI Mode, and LLM citations. Rebuilding measurement dashboards so browser-based traffic sits separately from total digital reach instead of mashed into one number.

This dataset is the starting line, not the finish. Overlay your business metrics on it and you'll see whether your audience is one of the ones moving or one of the ones still here. The Influenceable Web is contracting. The audiences inside it are still measurable, still receptive, and worth the work to reach.

From our own Momentic Research.

‍

Where the answer forms

If most of the recommendation is forming on surfaces you don't own, then most of the work is too. A VaynerMedia audit of one major brand found that the brand's own site was about 2% of the sources AI cited on unbranded questions about it. Eli Schwartz's seven-function framework points the same direction: an SEO team controls only two of the seven levers that shape answer-engine outcomes. Directional, single-brand, but it tracks with the rest of the data. Your site still matters. It stopped being where the answer gets made.

The surfaces

A handful of surfaces carry most of it: independent editorial is the largest share, communities like Reddit and forums are next, then third-party reviews and roundups. Your own site is the smallest piece. The work becomes: which specific publications, threads, and review surfaces does the model keep returning to for the questions your buyers ask, and how do you earn a place there.

What to distribute

You can't distribute nothing. The brands earning citations publish things a competitor cannot reproduce: first-party data, real testing, a clear point of view. Templated posts and AI-spun filler give a distribution engine no cargo, and they pull your brand into the same blur as everyone else writing the same thing. Scaling the right kind of work earns more presence. Scaling for its own sake quietly works against you.

Measurement

The clean attribution is gone. You will not tie a single AI mention to a single sale, and any tool that promises otherwise is selling you a story. Track whether you show up, how often, and whether the model describes you accurately or from something stale. Then prove value the way brand always has been proven: holdouts, correlations, brand-search lift over time.

The slow part

None of this is fast. Earning a position on the surfaces AI cites takes months, sometimes years, and that's the point. The brands willing to do the slow work end up owning a position competitors cannot buy back. Which is why I'm running this dataset monthly instead of waiting for an annual report. The shifts that decide who shows up next year are happening in the data now.

‍