I used to think of digital PR as the soft end of marketing, all placements you couldn't attribute and coverage reports nobody opened twice. Coming from performance marketing, a mention with no link looked like wasted money to me. Then one brand kept turning up in ChatGPT answers even though its rankings were mediocre and its backlink profile was thin. I went looking for the reason and found twelve trade-publication articles that named it without linking to it once.
You can put numbers on that now, and they aren't close. When Ahrefs studied 75,000 brands, web mentions correlated with AI visibility at 0.664, while backlinks came in at 0.218. So the signal SEO teams have chased for twenty years is three times weaker in AI answers than the one PR teams were producing all along. SEO is drifting into PR territory, and if your budget still treats links as the main currency, you're paying for the weaker of the two. The rest of this post is how I'd move that money.
The evidence that AI answers run on mentions
I know how this sounds coming from a PR agency, so start with how a model learns about you. It reads text: articles, roundups, forum threads, podcast transcripts. A link tells a crawler where to go next, while a mention tells the model something about your brand, and the model is the one writing your buyer's answer.
The Ahrefs study had one detail that should bother anyone who builds links for a living. Nofollow and dofollow links showed almost the same influence on AI visibility. The attribute Google trained a whole industry to fuss over makes no visible difference to the systems that now answer your buyers' questions. Rand Fishkin, who did a lot to build the link-obsessed era in the first place, has come around too:
Godfather of SEO, @randfish: “Mentions matter more than backlinks. If you’re still link building, I have great news, you can stop. You don’t have to do it. It’s like washing chicken, you don’t have to do it man, just stop. You can see where this train is going. Mentions.” Me: https://t.co/rt5JovMym3
Edward Sturm@edwardeachdayJul 7, 2026
Paraphrasing him in case the post disappears: mentions count for more than backlinks now, and chasing links for their own sake is on its way out. Search Engine Land's work on unlinked mentions adds a number that should change how you see your rankings. Nearly nine in ten pages ChatGPT cites sit outside the top 20 organic results, so the answer layer mostly ignores where you rank and reads the rest of the web.
That layer is growing fast, too. Wix's AI Search Lab data has AI search visits up 42.8% year over year, from 15.6 billion a quarter to 27.4 billion. More and more of your category's questions get answered there, so mentions stopped being a side channel a while ago.
The Ahrefs data holds a warning as well. A full 26% of the brands in it had no AI Overview presence at all, despite having websites, rankings and link profiles. Nobody had written about them in a way a model found worth repeating. We covered the backlink half of this in our own 20-site backlink test; the mention half is the one most teams still haven't budgeted for.
AI search traffic grew 835% this year. Is your content ready?
RankControl generates 26 content formats optimized for ChatGPT, Claude, and Perplexity. Published on your domain, matched to your brand.
Where mentions convert, and why not all coverage is equal
The most useful research I've found on which coverage converts didn't come from a PR firm. AirOps analyzed 21,311 brand mentions across ChatGPT, Claude and Perplexity, and 85% of them came from third-party sources. Put another way, a brand was 6.5x more likely to get named through somebody else's page than through its own.
Where that 85% lives is the part you can act on. Roughly 90% of those third-party mentions sat in listicles, comparisons and reviews, while feature profiles and funding announcements barely registered. The plain "best X for Y" roundup is the placement type that converts best in AI search. Our listicle strategy guide reached the same conclusion from the content side. So if a campaign could get you one profile in a business magazine or three slots in mid-tier category roundups, I'd take the roundups every time.
Two more findings should shape who you pitch. The first is that 68% of brand mentions were unique to a single AI model, so a strong showing on Perplexity tells you nothing about ChatGPT. Spreading placements across different kinds of publications hedges that.
The second comes from Wix's AI Search Lab, covered by Contently. Brands present on 4+ platforms had 2.8x higher odds of appearing in ChatGPT responses. A podcast with a transcript counts as a platform, and so do newsletters, trade blogs and communities, so a spread of smaller placements will do more for you than one big hit.
Does it work when someone aims it at AEO on purpose? One agency's documented campaign landed 151 placements across 92 outlets, and its daily Bing AI citations roughly doubled. That's one self-reported case, so I wouldn't lean on it alone, but it points the same way as the bigger studies.
The new PR brief: pitch for mention quality
What you ask for matters more than where you pitch. The old brief chased a followed link with whatever anchor text the editor allowed. A model reads the sentence your brand sits in, so build the new brief around that sentence. I'd ask for four things, in this order.
Start with your category in the same sentence as your name. "RankControl, an AI search visibility platform," teaches a model far more than the name dropped on its own, because models work out what you are from the words that keep showing up next to you. Seer Interactive's fan-out analysis shows brands that co-occur with category language surfacing across the sub-queries AI engines actually run.
After that, try to get a number attached to your name, because adjectives rarely survive into an answer. A line like "Customers using X cut reporting time 40%" hands the model a fact it can lift with your brand already inside it.
The third ask is the one that slips: the same facts everywhere. Your category label, your main claim and your numbers should read identically in every placement. Models trust a claim that repeats across independent sources, and when two sources disagree they treat both as noise. Write one canonical boilerplate and hold every outlet to it. Our Brand Control profile keeps that single description so every article RankControl writes uses it, since drift tends to sneak in one podcast intro at a time.
Last, ask for a table row whenever the format allows one. In Contently's reporting, comparison tables got extracted 81% of the time against 23% for prose. A row in someone else's comparison table is a structured fact about you on a domain the model already trusts.
A link isn't one of the four. Take it if it's offered, but don't trade any of the above for it.
For volume, the reactive channels still earn their keep. The HARO successors (Qwoted, Connectively, Featured) put you in front of journalists writing the listicles and roundups that models learn from. Expect single-digit win rates, and count each win as lasting text about you that models will keep reading.
Built by the team that got cited in 48 hours.
Content generation, backlink building, AI visibility tracking, and Google rankings. One platform, zero guesswork.
Mine the mentions you already have
Check what's already out there before you pay for anything new. If your brand is a few years old, the web almost certainly names it in places that never linked to you. The classic playbook treated those as link-reclamation targets: find the author, email them, plead for the link. For AEO you can skip most of that, because the mention already did the part that matters.
The workflow is short. Run your brand name through a content index; Ahrefs Content Explorer with your own domain excluded is what most people use. Filter down to pages with real authority and organic traffic, then read what each of those sentences says about you.
Fix what you find in a set order. Wrong facts come first, because a stale "starting at $99/mo" in a well-ranked roundup is exactly the kind of line a model repeats word for word. Ask for category framing next, and leave the link request for last as a bonus. Alerts will catch new mentions while they're fresh, but keep a quarterly re-audit anyway, because alerts on their own miss about a third of what a periodic sweep turns up.
We've seen this pattern more than once. A brand gets named in a handful of "best tools" threads and city-level listicles, with no links anywhere, and within weeks it starts appearing in AI answers for exactly those categories. The client always asks where the link is. There was never going to be one, and it never mattered, since the mention was doing the work the whole time.
Who owns this? (The gap where the value leaks)
The org chart might be a bigger problem than any tactic here. Earned mentions drive AI citations, yet mentions don't sit under SEO and AI citations don't sit under PR, so nobody owns the result. The people in this thread are running into that wall right now:
"Earned" brand mentions are driving AI citations but don't fall under SEO. What are teams doing about this?
Read a study that showed by a large margin, earned citations and social & community citations outweighed all on-page content factors. This was tracking 6 major AI models across 31 industries. (Question 1) How are people getting these off-si...
It matches what we hear from customers. PR lands the coverage, SEO reports that rankings didn't move, and the AI citations the campaign produced appear on nobody's dashboard. Google now reports impressions for linked URLs in AI features, but unlinked mentions, the currency that counts most here, show up nowhere. People running GEO programs keep pointing at that hole: the data you'd need to prove PR's AI value isn't in any Google product.
So you build the metric yourself. I call it mention-to-citation conversion. Log each placement as it lands, with the date, the outlet, the category framing and the exact sentence your brand sits in. Then run your category queries through ChatGPT, Perplexity, Gemini and Copilot every week. When your brand starts appearing in answers to the queries a placement aimed at, you've got a conversion you can attribute, which no coverage report ever gave you.
That loop is what our AI visibility tracking automates. Every tracked query gets checked on six engines every week, so a March placement that starts paying off in May shows up in the trend instead of being forgotten.
Doing it by hand? Count the recurring hours first, because that's where the cost hides:
| Task | Time |
|---|---|
| Initial unlinked-mention audit, established brand | 4-6 hours |
| Fact-correction outreach | 2-3 hours per batch |
| Weekly citation sampling, twenty queries on four engines | 3-4 hours, every week, indefinitely |
| Reactive pitching to keep new mentions coming | 2-3 hours weekly |
| Steady state, after a 10-hour setup | 6-8 hours a week |
Most founders run the first month with real energy. Month three is usually where the program dies, without anyone deciding to kill it.
Mentions also fade on their own. Models retrain, roundups get rewritten, and what the web said about you last year thins out. Earning the mention was the easier job; the harder one is noticing when the models stop repeating it, which a one-off campaign report won't show you.
See your first AI citation report in under 5 minutes.
No setup calls. No onboarding meetings. Connect your domain and see where AI mentions your brand right now.
Point the budget at words, not links
If you keep one idea from this post, keep this one: AI engines repeat what trusted text says about you. Your job is getting into that text with the right facts, in as many places as you can, and keeping those facts consistent. I'd begin by auditing the mentions you already have and fixing the facts in them. Then pitch roundups and comparisons with the new brief: category framing, a stat on your name, and no begging for links.
You can run the audits, the pitches, the boilerplate policing and the weekly citation sampling yourself. Or RankControl's agents can keep your brand facts consistent in every article they write, find pages that name you without a link each week, and track which answers cite you, while you get on with being worth mentioning.
