The New Search War: Why Companies Are Suddenly Paying to Be Remembered by Machines

A billion-dollar industry has formed around a question almost nobody was asking eighteen months ago: what does ChatGPT say about your company? Jake Vince, Co-Founder of S99 PR, on why the answer runs through the press rather than the search bar.

For most of the last two decades, the answer to "how do people find businesses" was simple enough that entire industries were built on it. People typed something into Google. Google returned ten blue links. Businesses fought over which of those links appeared first.That system is now competing with something structurally different, and the companies that noticed early are quietly rewriting their marketing budgets around it.

When someone opens ChatGPT and asks for the best commercial real estate attorney in Chicago, or the top three project management platforms for a fifty-person agency, they do not get ten links. They get an answer. Usually three to five specific companies, named directly, with a short explanation of why each one belongs on the list.There is no second page. There is no scrolling. For the businesses named, it is the most valuable placement in modern marketing. For everyone else, it is invisibility.

Jake Vince, Co-Founder of S99 PR, a Los Angeles public relations agency, says the change surfaced in client conversations long before it surfaced in any industry report. "We started hearing the same question on sales calls, and it came from people who had never asked us anything technical before," Vince says. "A founder would say, I asked ChatGPT about my own company and it either said nothing or it said something wrong. That question went from rare to constant in about six months."

The clearest signal of how fast this shifted is the money.

Profound, a startup that tracks how brands appear inside AI-generated answers, raised a $96 million Series C at a $1 billion valuation, bringing its total funding to roughly $155 million. The company did not exist in any meaningful form three years ago. It sells software that answers one question: is your brand showing up when AI systems respond to questions in your category?

It is not alone. A category of tools has emerged, priced from $29 a month to well into five figures, all built around monitoring, measuring, and attempting to influence what large language models say about a given company. The vocabulary arrived just as quickly. Answer Engine Optimization. Generative Engine Optimization. AI visibility. Machine relations. Most of these terms did not exist in 2024.

The underlying consumer behavior justifies the scramble. Research from Pew found that when Google displays an AI-generated summary at the top of a results page, users click a traditional search result on 8 percent of visits, compared to 15 percent when no summary appears. They click a source inside the AI summary itself roughly 1 percent of the time.

In other words, the AI answer is increasingly not a stop on the way to the destination. It is the destination.

Vince, whose agency S99 PR places editorial coverage for founders and executives, says the pattern held across industries. "It was founders, attorneys, medical practices, consumer brands. Completely different businesses arriving at the same realization in the same quarter. Nobody had a plan for it because nobody had needed one before."

The instinct for most companies has been to hand the problem to whoever already manages their search engine optimization. That instinct is understandable and mostly misplaced.

Research from Chatoptic in 2025 measured the correlation between where a page ranks on Google and whether ChatGPT recommends that company. The correlation came back at 0.034. Statistically, that is close to no relationship at all.

The reason is mechanical. Traditional search is a ranking system. It sorts pages against a query and returns them in order. Large language models are retrieval and synthesis systems. They assemble an answer from sources they consider credible, then explain that answer in their own words. A company can hold the top organic position on Google and still never surface in an AI response, because the model is not looking at rankings. It is looking at what independent sources have said.

That distinction has produced the single most consistent finding across the research so far.

An analysis by Ahrefs studied 75,000 brands to determine what actually predicted appearing in AI-generated answers. It was not backlink volume. It was not domain authority. It was not content output. The strongest predictor was branded web mentions, meaning references to a company published on properties the company does not own or control.

A separate analysis by Muck Rack examining more than a million AI prompts found that roughly 84 percent of what AI systems cite is earned editorial media. Paid and advertorial content accounted for approximately 0.3 percent."

That number reframes the entire conversation," says Vince of S99 PR. "Companies have spent years optimizing the one asset the model trusts least, which is their own website. A claim on your own site is a claim. The same claim reported by an independent publication starts to look like consensus. These systems are built to tell the difference."

The gap between how important this has become and how well it is understood has created something close to an unregulated market.

Agencies that rebranded from search optimization to AI visibility now quote retainers between $3,000 and $25,000 a month, often for work that consists largely of schema markup, technical audits, and restructured website copy. Independent testing has cast doubt on how much of that moves the needle. Ahrefs ran a controlled study tracking 1,885 pages that added structured data markup and found no significant change in AI citation rates. Google's own documentation states that no special schema is required for its AI features.

Measurement is equally unsettled. Large language models are probabilistic rather than deterministic, meaning the same question asked twice can return different companies, different sources, and different phrasing. Several monitoring platforms sample each query once per week and present the result as a ranking. Practitioners have started pushing back publicly, arguing that a single weekly sample of a system that varies by user, location, and moment is a snapshot rather than a trend.

"There is a lot of confident language in this space and not much of it is earned yet," says Vince, whose team at S99 PR runs monthly placement programs built around the problem. "We tell clients directly that anyone selling a guaranteed AI ranking is selling something that does not exist. What can be committed to is the input. You can consistently place a company into the sources these systems draw from, and you can measure the direction of travel honestly."

For all the noise, a few findings have been replicated across independent studies.

The most consequential involves distribution rather than content. A controlled study by Stacker and Scrunch measured citation rates for identical content in two locations. Published on a brand's own website, the material was cited in AI answers 8 percent of the time. Distributed through third-party news outlets, the same material reached a 34 percent citation rate. A 325 percent increase, driven entirely by where the content lived rather than what it said.

Recency has emerged as the second major factor, and it is tightening. Analysis of AI citation patterns found that roughly half of all citations now come from content published within the previous seven days. Nine months earlier, that midpoint sat at eighteen days. An analysis of 8,000 wire-distributed press releases found an average time to first AI citation of about eight hours, with releases cited within the first 24 hours going on to earn 15 times more citations over the following month than those that were not.

Format matters as well. Research from Evertune examining nearly 400 million AI citations across 25,000 URLs found that 63 percent pointed to list-format pages, with the large majority being ranked "best of" or "top" comparisons. Separate analysis found that roughly 44 percent of all cited material is extracted from the first 30 percent of a document, which suggests that answering the question in the opening lines matters more than anything that follows.

Taken together, the picture that emerges looks considerably more like public relations than like search engine optimization. Independent coverage, published frequently, in publications the models already trust, describing the same company in consistent terms.

Vince says the recency data is the finding that changed how S99 PR structures its work. "A single article used to be an asset you could point to for years. It still is for a human reading your website. But for these systems it fades in about a week. That is not a reason to stop doing press. It is a reason to stop treating it as a one-time purchase."

The measurement debate has obscured a simpler question, and the early numbers on it are unusually strong. Visitors arriving from AI platforms convert at roughly 4.4 times the rate of visitors from traditional organic search, according to analysis from Semrush.

Some datasets put the gap wider still. Webflow disclosed that approximately 8 percent of its signups now originate from AI chatbots, and that those users convert at six times the rate of ordinary organic traffic.

The reason is straightforward. A person clicking a Google result is beginning their evaluation. A person arriving from an AI answer has already been told, by a system they trust, that this company is a legitimate option. The recommendation did the qualifying.

Research from G2 in April 2026 found that 51 percent of B2B software buyers now begin product research with an AI assistant more often than with Google, and that 94 percent use one somewhere in the purchase process.

"The volume is still smaller than organic search, and anyone telling you otherwise is exaggerating," says Vince of S99 PR. "But the quality of it is not close. When ChatGPT names you, the trust question is already answered before the person reaches your site. We have clients who can tell you the exact week it started happening, because the sales conversations changed."

One failure mode has surprised practitioners, and it affects companies that have done nothing wrong.

Before a model will name a company or an individual, it has to be reasonably confident it knows who they are. It constructs that understanding by pulling names, titles, company affiliations, and locations from across the web and reconciling them into a single record. When those details conflict across a LinkedIn profile, a company site, a conference biography, and older press coverage, the model has three options. It can merge the entity with a similarly named one. It can default to whichever version appeared most often, which is frequently the oldest. Or it can decline to include the entity at all because confidence is too low.

The result is that a genuinely accomplished executive with inconsistent public information can be absent from an answer that names a less established competitor whose record is clean.

"That is the part clients find hardest to believe," says Vince of S99 PR. "You can have real coverage and still be invisible, because the system cannot resolve who you are. It is why we use the exact same description of a client in every piece we publish, month after month. Consistency is what builds the record."

The consensus forming among practitioners is not that search engine optimization is finished. Traditional ranking still influences Google's own AI summaries, and organic search still drives enormous volume.

The argument is narrower and more specific. The channel that determines whether a company gets named inside an AI answer runs through third-party publications rather than through owned properties, and it is governed by frequency rather than by one-time optimization.

That places a discipline built decades ago at the center of a problem that did not exist two years ago."The strange part is that the answer turned out to be the oldest thing in the industry," Vince says.

"Get real publications to write about you. Do it consistently. Say the same thing about who you are every time. That was good advice before any of this existed. It just happens to be the exact input these systems are built to reward."

S99 PR is a Los Angeles public relations agency specializing in media placements, digital credibility, and AI search visibility for founders, executives, and brands. The agency secures editorial coverage across more than 400 publications and works with immigration attorneys nationwide on USCIS-compliant press for O-1 and EB-1A extraordinary ability petitions. S99 PR was co-founded by Jake Vince and Chris Kolaskos and has published more than 4,500 articles for over 5,000 clients.

More information about S99 PR is available at www.s99pr.com.