Last Updated on June 22, 2026 by Becky Halls
So Google finally said the quiet part out loud…. Buying AI citations is a big no-no!
On 15 May 2026, Google updated its search spam policies to state – for the first time, in black and white – that trying to manipulate generative AI responses counts as spam. That includes the booming little cottage industry of “we’ll get your brand cited by ChatGPT” services, paid citation placements, and the various clever schemes people have been quietly testing to force their way into AI Overviews and AI Mode.
If you’ve spent the last year wondering whether the “buy your way into the AI answer” pitches landing in your inbox were too good to be true: yes. They were. And now they carry risk.
Let’s unpack what actually changed, why it matters, and what to do instead.
What Google actually said
The update was small in word count and large in consequence. Google extended its existing spam framework – the same one that’s penalised paid backlinks for years – to explicitly cover “attempts to manipulate generative AI responses in Google Search.” Buying or manipulating citations inside AI Mode and AI Overviews is now treated the same way as buying fake backlinks.
That’s the key mental model here. Google didn’t invent a new rule so much as point at the old rule and say this applies to the shiny new thing too. If a tactic would have got you a manual action in 2018, dressing it up in AI language doesn’t make it safe in 2026.
The timing isn’t random. AI Overviews now trigger on nearly 48% of all tracked queries, up roughly 58% year over year. When something touches half of all searches, a grey market forms around it fast – and Google has every incentive to protect the integrity of those answers before they lose user trust.
Why the “force a citation” market was always shaky
Here’s the thing the citation-selling crowd doesn’t put on the sales page: a lot of these tactics relied on poisoning the well. Research out of Cornell, covered by 404 Media, found that AI research agents pull user-generated content from sites like Reddit in roughly half of all queries – and that “a single poisoned Reddit comment can influence generated outputs for an entire cluster of related queries.”
Read that again. The reason buying citations worked in some cases is precisely the reason Google had to ban it: the systems are manipulable, and that manipulation degrades the answers everyone relies on.
Google’s own John Mueller has been waving this flag for a while. Back in August 2025 he warned on Bluesky that “the higher the urgency and the stronger the push of new acronyms, the more likely they’re just making spam and scamming.” If a vendor is selling you a four-letter acronym and a guarantee, that’s usually the tell.
As our founder puts it:
“The whole ‘pay us and we’ll get you cited by AI’ model was a backlink-buying scheme wearing a new hat. Google killing it isn’t bad news for honest marketers – it’s the best thing that could’ve happened. It means the brands that earn genuine authority finally stop competing against people who simply bought their way in.” Ian Naylor, 3way.social
What this means for link building (the legitimate kind)
Here’s the good news, and it’s genuinely good: this update strengthens the case for real link building rather than weakening it.
AI engines don’t conjure their citations from nowhere. They lean heavily on the open web – and on the same authority signals that have always mattered. Analysis of LLM citation behaviour consistently shows that domain authority, high-quality backlinks from strong sites, and a healthy count of unique referring domains are among the biggest drivers of whether you get cited. One finding: sites with over 32,000 referring domains are around 3.5x more likely to be cited by ChatGPT than sites with a couple hundred.
In other words, the route into AI answers is paved with the exact same bricks as the route to ranking well in classic search: real coverage, real mentions, real relationships, real authority. The shortcut just got closed off, which means the long road is now the only road – and that rewards anyone who’s been doing it properly all along.
“I tell clients to stop asking ‘how do I get cited by AI?’ and start asking ‘am I genuinely one of the best sources on this topic, and does the web reflect that?’ Earn the second one and make sure you’re doing everything right, and the citations follow. Try to fake the first one and you’re now playing a game Google has explicitly said it will penalise.”
What to actually do now
A few practical moves in light of the update:
Audit your recent “AI visibility” spend. If you’ve paid anyone for citation placements, forced mentions, or “LLM training” packages in the last year, treat that the way you’d treat a dodgy backlink supplier. Distance yourself before it becomes a liability. If you’re feeling overwhelmed with everything you need in place for AI in 2026 then choose someone that can help with AI visibility and guide you to ensure your site is setup in the best way possible – without using any dodgy tactics.
Double down on digital PR and earned mentions. Getting referenced by credible, relevant publications does double duty in 2026: it builds the backlink profile that helps classic rankings and feeds the authority signals AI engines use to pick sources.
Build genuine presence where AI looks. Reddit, LinkedIn, G2 and industry communities are heavily cited — but the operative word is genuine. Show up with real expertise, not promotional drops. Brands with profiles on review platforms like G2 see meaningfully higher citation odds, and that’s earned, not bought.
Make your content extractable. Clear structure, factual density, and quotable passages help AI engines lift your content cleanly. That’s optimisation, not manipulation – and it’s entirely within the rules.
Reframe success metrics. Track share of voice in AI answers over time, not whether you “got the citation” this week. It’s a slower, truer signal.
The headline here isn’t “Google ruined AI marketing.” It’s “Google just made the honest version of the game the only version worth playing.” For anyone building authority the right way, that’s a Tuesday well spent.
Buying AI Citations – FAQ
Is it now against Google’s rules to want to appear in AI Overviews?
No. Wanting to appear – and optimising legitimately to do so – is completely fine. What’s now classed as spam is manipulating or buying your way into those answers. Earning visibility through quality content and genuine authority is exactly what Google wants.
Will buying AI citations get my whole site penalised?
Google has folded this into its existing spam framework, the same one used for paid links. That framework can lead to manual actions and ranking suppression. There’s no public “AI citation penalty” counter, but treating it as a real risk is the sensible call.
How is buying AI citations different from digital PR?
Digital PR earns coverage by giving journalists and publishers something genuinely worth referencing. Buying AI citations pays for placement or artificially corroborates claims to trick the system. One creates real-world signals; the other fabricates them.
Do backlinks still matter for AI search?
Very much so. Backlinks and referring domains remain strong predictors of whether AI engines cite you, because LLMs draw on the open web and inherit its authority signals. Legitimate link building is arguably more valuable now, not less.
What about getting mentioned on Reddit – is that risky?
Participating authentically with real expertise is fine and genuinely helpful. Astroturfing, planting fake comments, or “poisoning” threads to game AI outputs is exactly the kind of manipulation now in scope. Intent and authenticity are the dividing line.
Should I cancel my AI-visibility tools?
Not necessarily – monitoring tools that measure your AI visibility are useful. The ones to be wary of are services that promise to force or buying AI citations. Measuring is fine; manipulating is not.
Written by the team at 3way.social, where we think link building should be earned, not engineered.
