Google’s own guide to showing up in AI search results has one blunt line for anyone paying for placements or outreach aimed at AI visibility:

“Seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem.” It reads like an argument against an entire category of marketing spend, including the kind BusySeed sells its clients. 

The line landed at a moment when brand mentions and citations inside ChatGPT, Gemini and Google’s own AI Overviews have become a real line item in B2B marketing budgets, the practice generally called generative engine optimization, not a side experiment.

Google’s advice, taken at face value, says stop. 

The distinction Google’s guidance skips 

“Google isn’t wrong that a pile of throwaway mentions does nothing said Omar Jenblat, founder and CEO of BusySeed, a B2B demand generation agency in New York.

“But ‘inauthentic mentions don’t help’ and ‘mentions don’t affect what gets cited’ are two different claims,” said Jenblat

“We track citation behavior across client accounts, and where a mention sits does measurable work. Google’s guidance doesn’t separate the two.,” he said.

The data behind that is not proprietary theory. It comes from a dataset of more than 137,000 tracked AI citations, and the pattern is consistent: position on the page predicts citation far more than the page’s general authority does.

Named first, an entity gets cited roughly 72 per cent of the time. Second or third, 62 per cent. Fourth, 52 per cent. Fifth to ninth, 36 per cent. Beyond fourteenth, it is close to noise. 

Freshness moves the same needle. Around 59 per cent of citations come from pages updated within the last three months, 80 per cent within six.

The date that counts is the last update, not the original publish date, which is a different signal than the authority Google’s guidance is defending. 

Where Google has a real point 

None of that rescues low-quality link buying. Company and vendor pages, not third-party placements, account for roughly 60% of citations in assistant answers to commercial queries, against about 26% from third-party listicles.

A brand with no page that plainly answers what a buyer is asking will not get named no matter how many mentions it buys.

That is the failure mode Google’s warning is actually describing, and BusySeed tells clients the same thing before recommending any paid placement. 

It is also why a disclosed sponsorship does not appear to block citation the way it blocks Google’s own link-graph credit.

DesignRush discloses sponsored influence in its own agency rankings and still ranks third organically for “best GEO agencies,” cited in AI Overviews thousands of times.

Search ranking and AI retrieval read different signals; treating a paid label as automatically disqualifying is its own kind of mistake. 

What this means for a marketing budget 

“The sequence matters more than the debate about authenticity,” Jenblat said. “Build the page that answers the actual buyer question first.

“Buy placements after that, on pages that get maintained, in a position that gets read” 

“Buying volume before you have anything worth citing is the exact spend Google is telling people to stop making, and they’d be right to stop making it.” he said.

BusySeed built Rankxa to track that gap on a schedule: what a company’s real AI competitive set looks like, who gets named for the questions that precede a shortlist, and which of a brand’s own pages the assistants already trust enough to cite.

Running that check once is useful. Running it monthly is what turns Google’s own citation data into a repeatable process instead of an argument, which is what  generative engine optimisation work actually is.

Google’s AI Overviews now reach more than 2.5 billion monthly users, while AI Mode has passed one billion, making citation visibility a commercial issue rather than a speculative SEO trend.