“What’s closing is the cheap part,” Shane Tepper told me. “The stretch where you can win position with work instead of budget.”
Tepper is co-founder of Resonate Labs, and I’d asked him a pointed question. Everyone in SEO talks about a closing window on earned AI visibility, but the surface area of AI search keeps expanding. More queries get answered by ChatGPT and Perplexity every month. More of those answers now carry clickable citations. So, what exactly is supposed to be closing?
His answer sent me back to a field I know firsthand, one that has nothing to do with chatbots and everything to do with what happens when a wide-open channel starts getting fenced off.
I was president of SEO-PR from 2003 to 2025, part of our early reputation was built on a tactic the industry called press release SEO, or “SEO PR,” and it worked because distributing an optimized release through a wire service stacked three separate payoffs at once: a direct ranking boost from keyword-rich anchor text pointing back to a client’s site, referral traffic from the link itself, and the indirect benefit when a journalist read the release and wrote an original story that linked back organically.
Then, on July 30, 2013, Google fenced off the first one. It updated its Link Schemes guidelines and explicitly classified optimized anchor text in press releases distributed on other sites as an unnatural link, the same category as paid advertising, on the reasoning that a company pays a wire service for that distribution rather than earning it editorially. Every client relying on that anchor text for direct SEO lift lost it overnight. What didn’t disappear was the referral traffic the distribution still generated, or the indirect benefit when a reporter picked up the story and linked to it on their own initiative. SEO-PR kept winning awards for years afterward, for clients including Rutgers University, because the agency adapted to what Google had actually fenced off instead of pretending the fence wasn’t there.
AI search is running a version of the same play, just faster, and I recognize the shape of it because I’ve watched a wide-open field get fenced off before. Tepper’s argument, and the one his data backs up, is that the fencing has already started, and the brands that establish themselves before it’s finished are the ones who’ll still have ground to stand on once it is.
The Window Isn’t A Date. It’s A Race Against Your Competitors
Tepper pointed me to an audit from Fuel Online that checked 1,000 enterprise domains. Sixty-two percent came back technically invisible to AI models. Ask those same brands a plain, unbranded question about their own category, the kind a buyer actually types, and the models fail to mention them 81% of the time.
That’s the size of the opening. Most of the field hasn’t shown up yet.
What closes it is speed, not scarcity. Profound’s data puts the median time to first citation for new content at 6.81 days. Get published, get retrieved, get cited, all inside a week. The only thing standing between a brand and that citation is how fast it moves. Tepper’s read is blunt: the timeline for the window closing isn’t fixed. It’s “however long it takes your competitors to wake up.” In crowded B2B categories, he said, that clock is already running.
The May 7 Spike Held, And It Split Cleanly By Category
I’d flagged the widely reported 157.7% jump in ChatGPT referral traffic on May 7 and asked Tepper whether it was a blip. It wasn’t. Similarweb called it a new baseline rather than a spike, and Profound tracked the same structural jump, roughly a doubling that stuck, across every brand basket it monitors. Three separate measurement methods landing on the same date is a strong tell that something changed inside OpenAI’s product that day, even though the company never announced it.
The category breakdown is the part worth sitting with if you sell into a buying committee. Profound found B2B software and SaaS brands saw daily referrals climb more than 200% above the pre-May 7 baseline. Financial services and fintech picked up roughly 60%. Ecommerce and retail barely moved, because product recommendations route through ChatGPT’s shopping surface rather than the branded-link flow that got the traffic bump. Categories where ChatGPT recommends a company gained. Categories where it recommends a product didn’t.
OpenAI And Perplexity Are Making Opposite Bets, And Both Are Rational
I asked Tepper whether it was premature to declare that paid AI placements will dominate, given that OpenAI is testing ads while Perplexity pulled its own back in February. He doesn’t read the split as confusion. He reads it as two companies with different economics making different bets. OpenAI has hundreds of millions of free users and an infrastructure bill that scales with them, so ads fund access, and self-serve advertising opened to any U.S. advertiser by May. Perplexity is smaller and is selling trust as the product, which is why it walked ads back and leaned into subscriptions instead, with one Perplexity executive telling the Financial Times that ads make users start doubting everything they see.
Here’s my own read on top of Tepper’s, and I’ll put it plainly: The more important signal is what OpenAI did in the same week it opened self-serve ads. It also started surfacing clickable branded links, on May 7. Embedding brand URLs and tracking which ones get clicked is precisely the click data an ad-ranking system needs to learn from. The platforms haven’t converged on a shared model for monetizing AI answers. But the one with the most users has quietly started building the plumbing for one, and that plumbing runs on the same organic click behavior that earned citations already generate. Ignore that at your own risk.
“AI Authority” Isn’t Links. It’s How Often You Get Named
I pushed Tepper on what “AI authority” concretely means, since it gets thrown around loosely. His answer: It’s how often a model retrieves you, trusts you, and reads you as current and specific enough to put your name inside the answer for the questions your buyers are actually asking. It is mostly not a training-data phenomenon, because the engines that matter for vendor research retrieve live rather than recite from a static index.
The strongest predictor, per Muck Rack’s analysis of more than 25 million AI-cited links, isn’t backlinks. It’s earned media mentions, which accounted for roughly 84% of citations, against 0.3% from paid placement. Measuring it well means abandoning the “did we show up once” test entirely. Tepper noted that the odds of getting the identical AI recommendation twice for the same prompt are under 1%, which means a single result is noise, not a position. The fix is to run a stable set of real buyer queries repeatedly across ChatGPT, Perplexity, and Google’s AI surfaces, and score for whether your brand gets named in the answer text itself, tracked as share of voice against named competitors over time.
Will This Erode The Way Organic Google Visibility Did? Probably, Partly
SEO practitioners have earned this scar tissue honestly. Most of them spent years building organic visibility on Google only to watch ads and AI Overviews chip away at it anyway. I asked Tepper directly why AI citations should play out any differently, and to his credit he didn’t dodge it. “They probably will erode, partly,” he said. Ads already show up in about a quarter of AI Overview results, up from roughly 5% a year ago.
I think that honesty is exactly why the earn-now argument holds up rather than falling apart. The case doesn’t rest on AI citations staying free forever. It rests on two things that are true today and won’t stay true indefinitely. First, most citations right now still come from earned work rather than ad spend, which means the entry price is low precisely because most brands haven’t bothered to pay it in effort. Second, even once the paid layer matures, the organic signal doesn’t zero out. It becomes the substrate the paid system trains on. The brands search already recommends organically are the ones whose click data will train the ad-ranking model, and the ones a buyer already half-recognizes when a sponsored answer shows up next to them. That’s exactly how paid search shook out. Advertisers who also ranked organically ended up paying less per click and converting better. The field got fenced, but the ones already standing on it kept the best ground.
What To Actually Do This Month
Tepper’s practitioner advice doesn’t require a budget line, which matters if you’re the one person on a marketing team who’s read this far and is wondering where to start.
First, run the audit yourself before you pay anyone to run it for you. Write down 15 to 20 questions a real buyer in your category would type into a chatbot, comparison questions, “best X for Y” questions, “how do I choose” questions, and ask them across ChatGPT, Perplexity, and Google’s AI Mode. Note where you appear, where a competitor appears instead, and where the field is wide open. That’s an afternoon of work, and it tells you your real starting position, not the one you assumed.
Second, chase mentions, not links. Because most off-page movement in AI answers comes from being talked about rather than from backlinks, the highest-leverage move is getting your brand into the third-party comparison articles, review platforms, and category roundups that these models actually retrieve from when they build an answer.
Third, check your robots.txt file today. A meaningful share of brands are technically invisible for one embarrassingly simple reason: they’re accidentally blocking GPTBot, ClaudeBot, or PerplexityBot from crawling their own site. Fixing that takes ten minutes and costs nothing.
One Number To Hold Loosely
The projection getting passed around most is eMarketer’s forecast of U.S. AI search ad spend climbing from about $1 billion in 2025 to $2.08 billion this year and up to $25.9 billion by 2029, which would put AI ads at roughly 13.6 percent of all U.S. search ad spending. It’s a real number from a credible house, but it’s a forecast built on assumptions about platform adoption and shoppable ad formats, not a measurement of anything that’s happened yet. eMarketer’s own analysts flag the obvious caveat: if AI answers keep suppressing clicks the way they’ve done in traditional search, that ad spend may not deliver the traffic the dollar figure implies, and Google may slow-walk full monetization until the economics are clearer. Truist, for comparison, models OpenAI’s ad revenue alone hitting $30 billion by 2030. Treat $26 billion as a credible midpoint, not a settled fact.
My Take
I’ve watched a wide-open field get fenced off before, and it always happens the same way. Google didn’t warn anyone before July 30, 2013. It updated a guidelines page, and optimized anchor text in press releases stopped counting for SEO overnight. What determined who kept winning after that fence went up wasn’t who complained loudest about it. It was who’d already built something the fence couldn’t take away, referral relationships, journalist relationships, results a client could point to. SEO-PR was still winning awards years after that guideline update because we’d built past the one tactic Google closed off. Tepper’s data says most brands still have time to build something that durable before this fence closes too. It also says that window has a size, not a shape, and it shrinks every week a competitor figures this out before you do.
More Resources:
- ChatGPT Ads: New Acquisition Channel Or Just Another Brand Tax?
- How AI Will Transform PR’s Role In SEO Strategy Over The Next 2 Years
- Google Says Search Is Fine, AI Insiders Say The Median Person Has No Future
Featured Image: Elenyska/Shutterstock
