PR is a search channel now

PR now decides whether an AI assistant knows your business exists. That is a change in what the channel is for, and most marketing teams have not restructured around it. Only about 40% of digital PR teams have any repeatable process for earning AI citations at all.

Start with why it matters commercially, because the case is stronger than the hype and also narrower.

The citation gap is worth real money

Seer Interactive tracked 53 brands, 5.47 million queries and 2.43 billion impressions from January 2025 to February 2026. When an AI Overview appears on a search result, organic click-through sits at roughly 2.1% if your page is cited inside that overview, and roughly 0.9% if it is not. On a result with no AI Overview at all, it is about 3.3%.

Same query. Same ranking. Being cited or not is a bit over double the clicks.

Meanwhile zero-click searches reached 68.01% of US Google searches between January and April 2026, up from 60.45% in 2024. The click did not vanish. It got rationed, and citation is the ration card.

The second number worth knowing: traffic that arrives from AI assistants converts far better than search traffic. Semrush puts AI-referred visitors at around 4.4 times the conversion rate of organic. Seer's client data shows ChatGPT referrals converting near 16% against about 1.8% for Google organic. The volume is still small, usually 1% of sessions or less, and it grew 206% year on year from January 2025 to January 2026. Small, growing fast, and the highest-intent traffic on the site.

What Google actually published

In May 2026 Google shipped a guide called Optimising for generative AI features. Read it before you buy anything sold as AI visibility, because it kills several claims outright.

Google says its Search does not use llms.txt files. They neither help nor hurt.

Google says structured data is not required for generative AI search.

Google says there is no requirement to break your content into small chunks.

And, most relevant here, Google says manufacturing brand mentions across the web is not as helpful as it sounds.

That last line looks like it contradicts the whole premise, so sit with it. The correlation studies are real: brand mentions across credible sites track with AI visibility more closely than backlinks do. Google's statement is about manufactured mentions, the syndicated-quote farms and the paid listicle inclusions that exist only to be counted. Those are the ones that do not work.

The honest reading is this. Mentions correlate with visibility because both are downstream of the same thing: a business that genuinely gets written about by people who were not paid to write about it. You cannot fake your way up the correlation, and the tactics sold to do exactly that are the tactics Google named.

What earns a citation

The mechanism is retrieval. Assistants run what Google calls query fan-out, firing several related queries at once, then ground their answer in what comes back. So your job is to be the source that answers a specific sub-question better than anything else retrieved.

That means:

Be indexable and snippet-eligible. Google's guide is explicit that the page must be indexed and eligible to appear with a snippet. If someone put nosnippet on your templates for scraping reasons, they turned off your AI eligibility too. Check this first, before any content work.

Publish the thing nobody else has. Google's own worked example of what not to publish is a piece called something like 7 Tips for First-Time Homebuyers. That listicle format, the format still being sold as content marketing, is named in Google's documentation as commodity content. Original data, first-hand experience, specifics that only an operator would know: those get retrieved because nothing else answers the query.

Get quoted in places that get retrieved. Industry press, trade publications, association journals, university and government pages, and the specialist sites your buyers read. One quote in a trade publication that ranks for a topic is worth more than twenty syndicated releases.

Be consistent about who you are. Same legal entity name, same descriptors, same executive titles, same category language, everywhere. Assistants assemble an entity from fragments. If your chief executive is a Director on one page and a Founder on another and a Managing Director on LinkedIn, you have made yourself harder to assemble.

How to measure it without buying another platform

Pick 25 questions a buyer would genuinely ask before shortlisting a supplier in your category. Not your brand name. The category questions. Run them monthly in ChatGPT, Gemini, Perplexity and Google AI Mode, in a clean session, and record three things: whether you appear, whether you are cited with a link, and which other sources are cited instead.

That last column is the useful one. The sites cited in your category are your PR target list, ranked by evidence rather than by media kit.

Then set your referral tracking to separate AI sources properly, because default channel groupings bury them. Assistant referrals are worth watching at the session level even at 1% of traffic, because of the conversion rate.

The part that is uncomfortable

AI Mode itself is still 0.34% of searches. The strategic panic about it is running about a hundred times ahead of its actual query share. AI Overviews inside normal search results is where the traffic change is happening, and that is a different surface with different behaviour.

So do not restructure your entire marketing function around a surface with a third of one percent of queries. Do fix snippet eligibility this week, do stop commissioning commodity listicles, and do rebuild the PR target list around the sites that actually get retrieved in your category.

The channel that used to be measured in coverage is now measured in whether a machine repeats you when nobody typed your name.