The sections below distinguish confirmed Google guidance from findings in third-party citation research.
1. What Google’s Open Knowledge Format means for website owners
Google Cloud’s Open Knowledge Format (OKF) v0.2 adds optional fields for sources, verification, freshness and lifecycle status to knowledge bundles used by AI agents. Google has not connected those fields to Search rankings or AI citations, and businesses do not need to implement OKF to appear in AI Overviews or AI Mode. The useful takeaway is operational: keep important claims traceable, business information current and responsibility clear.
2. E-E-A-T remains a useful quality framework
E-E-A-T—experience, expertise, authoritativeness and trustworthiness—remains part of Google’s guidance for creating helpful, reliable, people-first content. Google also encourages publishers to make it clear who created content when readers would reasonably expect that information.
However, Google does not describe E-E-A-T as a single score or a pass-or-fail filter. It is therefore more accurate to treat it as a quality framework than to claim that weak E-E-A-T costs a specific number of ranking positions or automatically excludes a page from an AI-generated answer.
In practice:
- Use an accurate byline and author biography when the creator’s identity and experience matter.
- Demonstrate first-hand experience through examples, screenshots, tests, original data or a clearly described process.
- Support consequential claims with reliable sources.
- Correct substantive errors openly.
- Keep company, contact and service information accurate and consistent.
3. What small citation studies suggest about answer placement
Two independent 2026 studies examined where cited passages appeared within sampled webpages. Their findings provide useful directional evidence, but they do not establish a universal Google ranking or retrieval rule. These studies show patterns in their samples; they do not establish a ranking factor or guarantee citations.
- In CXL’s study of 100 Google AI Overview citations, 55% came from the first 30% of a page, while 21% came from the bottom 40%. This suggests that concise early explanations may be easier to retrieve in some contexts; it does not mean Google ignores the rest of a page.
- In Kevin Indig’s analysis of 18,012 verified ChatGPT citations, 44.2% came from the first 30% of the source text. Indig described the distribution as a “ski ramp” because citation frequency declined after the first third of the document. This supports making important information easy to locate, but it should not be treated as a guaranteed citation formula.
CXL also observed citations from self-contained question-and-answer sections. That supports writing clear answers when FAQs genuinely help readers, but it does not make FAQ sections or FAQ schema a guaranteed citation tactic.
The practical recommendation is simple: give readers a direct answer early when doing so feels natural, then provide context, evidence and nuance. Google explicitly says there is no requirement to “chunk” content into small pieces or rewrite it specifically for generative AI.
4. Independent corroboration can strengthen credibility
Consistent third-party information can help people evaluate a business. Reviews, reputable coverage, professional profiles and relevant forum discussions may provide context that a company’s own marketing cannot.
That does not mean corroboration has replaced backlinks, or that Google has confirmed a simple cross-platform agreement signal for AI citations. Links, mentions, reviews and business information serve different purposes.
The practical goal is accuracy. Make sure important facts—such as services, credentials, locations and pricing—are truthful and consistent wherever they appear. Work toward genuine reviews and relevant coverage, but do not manufacture mentions. Google’s guidance specifically warns that seeking inauthentic mentions is not a shortcut to visibility in generative Search. Google has also introduced user-facing features such as About This Result and Perspectives to give searchers more context; their existence does not show that they determine which pages Google’s AI cites.
5. What does not matter as much as the hype suggests
Two frequently recommended tactics deserve more scrutiny.
llms.txt. For Google Search, an llms.txt file is unnecessary. Google says that Search ignores llms.txt and other AI-specific files, meaning they neither improve nor harm Google rankings or generative-search visibility. Other services may choose to use the file, but most businesses should not prioritize it ahead of crawlability, useful content, internal links and accurate business information.
Structured data as an AI ranking tactic. Structured data remains useful when it accurately describes visible page content and supports an eligible Google search feature. For example, Article structured data can help Google understand an article’s title, images, dates and author information.
However, Google says that structured data is not required for generative AI Search and that there is no special AI schema. Its broader structured-data documentation describes markup as a way to give Google explicit information about a page and establish eligibility for supported search features. Treat it as accurate labeling, not as a ranking button or proven citation multiplier.
Five trust checks for an independent service business
- Confirm the basics: Make services, locations or service areas, contact details and hours accurate.
- Show who is responsible: Use an accurate byline, explain relevant team expertise, and include credentials only when they apply.
- Add evidence to service pages: Explain the process and what customers can expect. Use original examples, photos or sourced facts where available.
- Keep third-party information genuine: Correct inaccurate listings, respond to reviews and earn relevant coverage. Don’t manufacture mentions.
- Review and measure: Correct stale information, check Search Console and track whether pages support calls, bookings or qualified inquiries.