A Semrush study found that 80% of Reddit threads cited by AI systems have fewer than 20 upvotes, and most cited posts are roughly 900 days old — evidence that AI models reward genuine, aged human signals rather than engineered ones. A new analysis by Noah Malamut of PR firm LIMELIGHT argues this data point exposes the core weakness in the fast-growing Generative Engine Optimization (GEO) industry: the universal playbook of structural tricks has already stopped differentiating brands.

Why the standard GEO playbook no longer works

Search for GEO guidance and you find article titles like "Mastering Generative Engine Optimization in 2026" and "The Complete 2026 Guide of GEO for B2B." Malamut's observation is blunt: strip away the AI vocabulary and most recycle the same SEO playbook. In the SEO era, brands competed on a fixed set of mechanisms — keyword placement, backlinks, site speed, schema. GEO advice today mirrors that convergence: FAQ blocks, front-loaded answers, descriptive headers.

None of that advice is wrong, Malamut writes. But once every brand adopts the same structural tricks, those tricks stop being differentiators. "They are merely (important) maintenance practices that all brands will need to conduct," he said. Formatting rules become the new eligibility floor — necessary service, not competitive advantage.

Does manufacturing authenticity work?

Some GEO practitioners create fake Reddit threads to engineer authenticity, a tactic that looks coherent on its face because AI models cite Reddit more than any other platform. The Semrush data cuts against it: the threads models actually cite show indicators of human traffic and unfiltered opinion — low upvote counts from real communities and post ages averaging around 900 days, signals that are expensive or impossible to fake convincingly.

A separate California Management Review analysis reinforces the point. Because any single signal is now fakeable, audiences have stopped trusting isolated cues and instead evaluate a coherent picture of credibility, transparency, and reputation together. When all three align, the review found, the result is a positive trust outcome 82% of the time.

The conclusion Malamut draws: manufacturing authenticity harms the very brand reputation it is meant to build. "There is no way to game the system," he writes.

What separates durable AI visibility from tactics

The analysis frames the split between SEO and GEO precisely. Structural fixes correctly help models parse content — clear headers, answer-first formatting — but they only keep a site eligible. The durable differentiators, per the analysis, are:

• Earned media, including trusted-outlet bylines
• Original, proprietary research that other sources cite
• Genuine expert positioning and analyst quotes
• Peer-reviewed citations and third-party credibility built continuously

"When everybody has the newest technical fixes, it's the brands that have built rock-solid credibility that will surface first," Malamut writes. Algorithms will keep changing how they weight and surface signals, but they cannot manufacture a Wall Street Journal byline, a genuine analyst quote, or a peer-reviewed citation out of thin air.

Impact by site type and vertical

For B2B brands — the analysis's primary focus — the implication is that a serious GEO strategy must budget for both structural upkeep and continuous third-party credibility building, not one or the other. Publishers and brands already investing in original research and expert commentary hold a compounding advantage: every genuine point of view quoted in a trusted outlet strengthens the authenticity signal that models already favor.

The 900-day-old Reddit threads written by real users possessed an authenticity that cannot be manufactured, Malamut argues — and brands that built real earned authority will carry a self-magnifying advantage into whatever the next shift in AI search brings, while others scramble for "How to Master GEO in 2027" articles.

What to monitor next

Watch how AI systems evolve their citation patterns as the 2026 GEO guides proliferate — if structural convergence continues, expect earned media, original research citations, and verifiable expert attribution to carry increasing weight as the tiebreakers in AI-generated answers.