Cited Newsletter Issue 21 (May 15th 2026)

Welcome back to the Cited newsletter, your fortnightly briefing on what's actually moving in AI search visibility. Issue 21 covers a lot of ground: a significant shift in how Perplexity is weighting sources, fresh data from our own platform on citation velocity, and a frank look at why some brands that invested heavily in GEO early are now seeing their rankings stall. We've also got a short breakdown of what's working in structured content right now, and a few signals worth watching into Q3 2026.

If you're reading this for the first time, the Cited newsletter is published by the team at Lua Rank. We build AI visibility tools, run optimisation programmes for 40+ brands, and share what we're learning week to week. No filler, no vendor hype. Just what we're seeing.

What's Changed in AI Search This Fortnight

Perplexity's Source Weighting Has Shifted Again

This one caught several of our clients off guard. Around May 8th, multiple brands tracking their Perplexity citations through Lua noticed a meaningful drop in citation frequency for content that had been performing consistently for six to eight weeks. After digging into the pattern, the picture became clearer.

Perplexity appears to be placing more weight on recency signals and topical clustering rather than single-page authority alone. Brands that had one or two well-optimised answer pages were hit harder than brands maintaining a consistent publishing cadence across a topical cluster. This is not entirely new behaviour, but the weighting shift is more pronounced than anything we've tracked in the previous four months.

The practical implication: if you built your Perplexity visibility on a handful of cornerstone pages and then slowed your content output, now is the time to reactivate your publishing schedule. A cluster of three to five supporting articles around your core topics, published over two to three weeks, tends to restore and often improve citation rates.

Google AI Overviews: The Attribution Gap Is Narrowing

One of the persistent frustrations with Google AI Overviews has been the inconsistency between which sources are cited visibly and which sources actually inform the answer. Independent research from the Search Engine Journal suggests that visible attribution in AI Overviews correlates more strongly with structured data implementation than with domain authority metrics traditionally used in SEO ranking models. Source: Search Engine Journal

We're seeing this in our own data. Brands that completed Lua's structured data layer (one of the 13 optimisation layers we assess) are appearing in AI Overview citations at a rate roughly 2.3x higher than brands that haven't touched their schema markup. That's across a sample of 38 tracked domains over the past 60 days.

What "structured data implementation" actually means here

We're not talking about basic article schema. The meaningful lift is coming from:

• FAQPage schema with genuinely useful question/answer pairs (not thin padding)

• HowTo schema on process-oriented content

• Speakable schema, which remains underused and appears to carry disproportionate weight in AI Overview selection

• Entity disambiguation via Organisation and Person schema connected to Wikidata identifiers

If your content team hasn't prioritised this, the window for competitive advantage here is still open, but it's narrowing. More brands are catching on.

Data From the Lua Platform: Citation Velocity Trends in May 2026

Every two weeks we pull aggregate data from across the brands running active programmes on Lua. Here's what May 2026 looks like so far.

The Perplexity slowdown is consistent with the source weighting shift we described above. The Google AI Overviews acceleration is notable and tracks with the structured data findings. Claude continues to reward depth over frequency, which makes it harder to move quickly but more defensible once you're cited.

The Early Mover Stall: Why Some Brands Are Plateauing

We've started seeing a pattern worth naming. Brands that moved fast on GEO in late 2024 and early 2025 built initial citation positions relatively quickly. Now, roughly 12 to 18 months later, some of those same brands are flattening or declining, while newer entrants are catching up.

There's a fair counterargument here: some teams would say that early mover advantage in AI search is durable because AI models are conservative about changing trusted sources. There's truth in that. But it's not unconditional. The brands plateauing share a common pattern: they stopped executing. They did a burst of optimisation work, got results, and then treated it as "done." AI models don't treat your content as permanently authoritative. They re-evaluate sources continuously based on freshness, accuracy, and competitive signal strength.

Consistency of execution is the moat, not the initial sprint.

What to Watch in Q3 2026

Multimodal Citations Are Coming

Both Google and OpenAI have been expanding their multimodal capabilities throughout 2025 and into 2026. The next frontier for AI visibility isn't just text citations. It's whether your brand's visual content, video transcripts, and audio content get extracted and attributed in multimodal AI responses. Research from the Reuters Institute's 2025 Digital News Report points to audio and video as rapidly growing AI-sourced content formats. Source: Reuters Institute

We're building multimodal tracking into Lua's roadmap for Q3. In the meantime, brands that maintain transcribed video content and structured audio pages are better positioned than those who aren't.

Regulatory Pressure on AI Citation Attribution

The EU AI Act's provisions around transparency in AI-generated content are pushing model providers toward more explicit source attribution. This is a slow-moving change, but directionally it favours brands that have invested in building genuine topical authority rather than those trying to game citation patterns with thin content. Source: EU AI Act, Official Journal of the European Union

Longer term, this regulatory trajectory could make AI citation visibility a more auditable and standardised metric, which is good news for brands that have been doing the work properly.

The Agent Search Layer

One signal we're watching closely: AI agents that conduct multi-step research tasks on behalf of users are becoming more widely used. When an agent researches a topic, it tends to favour sources that return consistent, structured, extractable answers across multiple queries. This is different from single-query citation optimisation. Brands optimised for agent search will need content that holds up under repeated, varied interrogation, not just one well-crafted answer page.

Gartner's 2025 research on AI search adoption suggests that agentic AI use cases will represent a significant share of enterprise search behaviour by 2027. Source: Gartner We think the brands building structured, consistent content programmes now are already preparing for this shift, whether they realise it or not.

That's Issue 21 of the Cited newsletter. If you're not already tracking your brand's citation performance across ChatGPT, Perplexity, Google AI Overviews, and Claude, Lua is the fastest way to get a clear picture of where you stand and what to do about it. We'll be back on May 29th with the next edition.