New Research Outlines 23 Key Factors for Content to Be Cited in AI Search Engines

If you’ve been wondering whether AI search engines require a completely new visibility playbook, the latest research points to a more balanced answer: the fundamentals still matter, but content format and extractability matter more than ever.

A new analysis from Cyrus Shepard, published through Zyppy Signal and quickly picked up across the SEO and AI search community, pulls together findings from 54 studies, experiments, patents, and case studies. The result is a list of 23 factors correlated with citation visibility in AI-powered search experiences, including Google AI Overviews, Gemini, Perplexity, and ChatGPT.

The clearest takeaway is this: getting cited by AI isn’t random, and it isn’t separate from SEO. Traditional search strength and AI citation performance appear to be closely linked.

Why this research matters right now

Over the past two years, search has moved sharply toward generative answers. Instead of sending users to a page of blue links, platforms now summarize, synthesize, and cite selected sources. That shift has serious implications for publishers and brands.

The data behind it is hard to ignore. Studies referenced in the research point to steep click-through declines in traditional results when AI answers appear. At the same time, pages that do earn citations can gain outsized visibility and clicks. Being left out of AI answers is becoming expensive; being included can be unusually valuable.

That’s why Shepard’s work is drawing attention. Rather than relying on hunches, it brings together a broad evidence base and turns it into a practical framework for what makes content citable.

The core finding: AI citation rewards SEO plus extractability

One of the most useful points in the research is that these are not classic ranking factors in the old-school sense. They’re correlated features, signals that repeatedly appear in cited content across multiple studies and platforms.

Even so, the patterns are strong enough to shape strategy.

At the top of the list are factors that will feel familiar to anyone in SEO:

• URL accessibility
• Strong search rankings
• Topic cluster and fan-out query relevance
• Proper preview controls
• Page-level authority and trust

That suggests AI engines are not replacing the web’s existing authority systems. They’re building on top of them. If your page can’t be crawled, can’t be parsed, or doesn’t rank well for relevant queries, it becomes much less likely to appear in AI-generated answers.

But AI search adds a new requirement: your content also needs to be easy for models to extract, verify, and quote.

The 23-factor model changes how content should be written

What stands out most is how strongly AI systems seem to favor content structured for direct retrieval and reuse.

Among the highest-value traits are:

• Answers placed near the top of the page
• AI-ready structure
• Self-contained passages
• Explicit phrasing
• Factual specificity
• Clear, chunkable formatting

That fits the way retrieval-augmented generation works. AI systems often pull from multiple sources, compare passages, and assemble responses from the most trustworthy and extractable material. If your best answer is buried halfway down the page, wrapped in vague language, or dependent on surrounding context, it becomes harder to cite.

That’s a meaningful strategic shift. For years, many brands optimized for long engagement, storytelling, and gradual information delivery. Those approaches still matter for human readers, but AI systems seem to reward content that gets to the point quickly and cleanly.

What the strongest factors appear to be

Based on the synthesized scoring, the most influential factors include accessibility, rankings, preview permissions, and topical depth. In practical terms, that points to a few clear priorities.

First, your content needs to be available to the systems that power AI answers. If you heavily restrict crawling or snippet generation, you may lower your chances of being cited. That trade-off is one of the more important points in the research. Many publishers want to limit AI usage, but the same controls can also reduce discoverability in AI-driven interfaces.

Second, topical authority matters more than isolated pages. The emphasis on fan-out queries and topic cluster rank is especially revealing. AI systems don’t just evaluate one page at a time; they often move through related sub-questions. Brands with strong supporting content across a topic may be better positioned to surface in these broader retrieval paths.

Third, directness wins. Pages that clearly answer likely user questions, use explicit language, and present verifiable facts appear to have a structural advantage.

Some signals matter less than many expected

Another helpful part of the research is what it pushes down the list.

A few widely discussed tactics appear to have weaker or less consistent impact, including:

• LLMs.txt
• Domain authority as a standalone metric
• Structured data on its own

That doesn’t make them useless. It means they’re not magic switches for AI citation. For marketers chasing shortcuts, that’s an important correction. AI visibility seems to be earned more through content quality, page accessibility, and topical alignment than through isolated technical add-ons.

Different AI platforms may behave differently

The research also highlights an important nuance: not all AI engines retrieve and cite content the same way.

ChatGPT and Perplexity may rely more heavily on known-source effects and established authority signals, while Gemini appears more retrieval-focused. So brands shouldn’t assume one-size-fits-all optimization. There is overlap, but platform behavior still varies.

Even so, the overall direction is consistent. Whether a system pulls from live retrieval, grounded search results, or a blend of training and search data, the pages most likely to be cited tend to be relevant, trusted, well-structured, and easy to extract from.

What marketers should do next

If you were turning this research into an action plan, five priorities stand out.

1. Strengthen traditional rankings

The evidence still points to Google visibility as a major predictor of AI citation. Strong rankings remain foundational.

2. Put the answer early

Lead with the clearest, most useful answer near the top of the page. Don’t make AI systems work to find it.

3. Write in self-contained blocks

Create passages that make sense on their own. This improves extractability and makes citation more likely.

4. Build topical clusters

Cover the main query, adjacent subtopics, and likely follow-up questions. AI engines increasingly evaluate topics, not just pages.

5. Avoid blocking yourself out

Review preview controls, snippet settings, and crawler rules carefully. Overprotection can cost visibility.

The bigger picture: SEO and AI visibility are converging

The most encouraging part of this research is that it doesn’t suggest marketers need to throw out SEO and start over. Instead, it shows that the future belongs to content that works for both humans and machines.

Authority still matters. Relevance still matters. Trust still matters. But now formatting, chunking, clarity, and citation-readiness matter too.

The web is moving toward a citation-first model, where the best-performing content isn’t just discoverable, but quotable.

FAQ

Does AI citation require a completely new SEO strategy?

No. The research suggests AI citation builds on traditional SEO rather than replacing it. Strong rankings, crawlability, and authority still matter, but content also needs to be easier to extract and quote.

What kind of content is most likely to be cited by AI search engines?

Content that answers questions early, uses explicit language, includes verifiable facts, and is organized into clear, self-contained sections appears to have the best chance.

Are technical tactics like structured data enough on their own?

No. Structured data can still help, but it does not appear to be a standalone driver of AI citation. Content quality, accessibility, and topical relevance carry more weight.

Do all AI search platforms behave the same way?

No. The research suggests there are platform differences. Some systems may lean more on authority signals, while others appear more retrieval-driven.

Conclusion

This 23-factor framework gives marketers something the industry has needed: a more evidence-based way to think about AI citation. The lesson is straightforward—winning in AI search means combining strong SEO fundamentals with content designed to be extracted, trusted, and cited. If you’re serious about building content for that reality, using tools built for it can help, and AIuthority is a strong place to start.