AI answers are built on what others say about you. PR is back on the agenda.

For years, earned media was the line item that was hardest to defend in a budget meeting. This week's news suggests it is about to become one of the easiest.

What happened

AI chatbots lean on third-party coverage when they answer questions and recommend brands: reviews, comparison guides and articles written by someone other than the brand. One forecast projects that PR and earned media budgets will double by 2027 as more people ask chatbots instead of searching. It is a forecast, not a fact, but the direction matches everything else in the news.

Alphabet's quarter is part of that picture. Google's advertising business missed analysts' estimates, and growth in search advertising revenue slowed, while its cloud business beat expectations with $30.7 billion of Alphabet's $127.6 billion in sales. Google is still working out how to turn AI into faster growth in search ads.

Then there is a useful piece of research. Semrush combined its AI citation data with LinkedIn's engagement data and analyzed around 89,000 LinkedIn URLs cited by AI search tools. 95% of the cited posts were original content rather than reshares, and most of them averaged only 15 to 25 reactions.

Meanwhile the feeds keep moving away from who you follow. Reddit redesigned its feed to push people beyond the communities they already joined, and TikTok launched restaurant guides built from more than 40 million posts and comments.

What it means if you lead marketing

The AI answer about your brand is assembled from what other people have written. Your own site still matters, but the reviews, comparisons and independent articles carry more weight than they used to. That puts PR, analyst relations and customer reviews back at the center of the plan.

The Semrush finding is the part I find most encouraging. The posts that AI tools cite are not the ones with thousands of likes. They are original posts with modest engagement. That means a steady habit of publishing real, first-hand thinking can earn visibility without a viral moment. Reach and being cited are becoming two different scorecards.

Two cautions. First, one argument this week made the point that AI answers mostly reach buyers who are already in the market. They capture demand, they don't create it. Judge that work like brand and consideration, not like a performance channel. Second, another piece noted that being named by an AI is easy and being described correctly takes work. Check what the assistants actually say about you, not just whether you appear.

What I would do

• Ask the main AI assistants about your category once a month. Write down whether you appear, how you are described, and which sources they lean on.
• Give earned media a real owner and a real budget line. Treat independent reviews and comparison coverage as a program, not a lucky break.
• Publish original thinking on a schedule. First-hand posts from your leaders, even with modest engagement, are what AI tools tend to cite.
• Fix the description, not just the mention. If an assistant gets your product wrong, find the sources it is reading and correct them at the source.

Worth trying

A third-party citation scoring worksheet. One guide this week shares a framework that scores the websites influencing AI recommendations across six criteria, including citation influence, brand association and business value, and it comes with a worksheet and a Claude skill. I like the discipline of it: instead of chasing every mention, you rank which independent reviews and comparison pages actually matter, and work those first.

The budget conversation is changing. The brands that invest in being talked about, accurately and by others, will show up in the answers.