Google publishes guide to optimising for generative AI search

Google has released an official guide to help publishers and content creators optimise their websites for generative AI search results. The guide provides best practices for ensuring your content appears prominently in AI-powered search features. Learn the key strategies to improve your digital visibility in this new search landscape.

Google's new guide to optimising for generative AI search, announced by John Mueller on 15th May 2026, marks the first time the company has consolidated its scattered statements about AI Overviews into official Search Central documentation. The 15-page document confirms what many practitioners suspected: AI search visibility isn't a separate discipline requiring entirely new tactics—it's an evolution of existing SEO, powered by the same core ranking systems you're already optimising for.

But here's what Google's guide doesn't tell you: how to audit your existing content for AI-readiness, which of your pages are being cited in AI Overviews right now, how to measure your share of AI visibility against competitors, and how UK businesses should adapt these principles to local search behaviour. This article fills those gaps with a practical framework you can implement immediately.

Key Takeaways

• AI search uses your existing Search rankings: Google's generative AI features pull from the same index and ranking systems as traditional search results—strong SEO fundamentals remain your foundation.
• Retrieval-augmented generation (RAG) is the mechanism: Google's AI doesn't "know" your content—it retrieves relevant pages from Search, then generates responses based on what it finds.
• LLM-specific tactics don't work: Google explicitly states you shouldn't create separate LLM-optimised content, use special markup, or attempt to "prompt engineer" your pages.
• Structured data remains valuable: Existing schema markup (especially for FAQs, How-tos, Products, and Reviews) helps Google understand and surface your content in AI features.
• Agentic experiences are coming: Google's guide introduces the concept of AI taking actions on users' behalf—booking appointments, making purchases—requiring new technical preparation.
• Measurement is the missing piece: Google's guide doesn't explain how to track AI visibility—you'll need third-party tools and indirect signals to understand your performance.
• UK businesses face unique challenges: Local search behaviour, NHS and government content prioritisation, and GDPR considerations create specific optimisation requirements the guide doesn't address.
• Content depth beats keyword optimisation: AI features favour comprehensive, authoritative content that answers follow-up questions—not keyword-stuffed pages targeting single queries.

Why Google Published This Guide Now: The Strategic Context

Google's decision to publish official documentation in May 2026 wasn't arbitrary. AI Overviews had been rolling out globally since May 2024, and the company had faced mounting criticism for two years about the opacity of its AI search systems. Publishers reported traffic declines they couldn't explain. SEO practitioners shared conflicting theories about optimisation tactics. Industry conferences featured Google representatives giving vague reassurances without actionable guidance.

The guide represents Google's attempt to establish an official narrative before misinformation became entrenched industry wisdom. By consolidating scattered statements from blog posts, conference presentations, and interviews into formal documentation, Google aims to prevent the kind of counterproductive optimisation tactics that plagued early SEO—keyword stuffing, but for AI.

For UK businesses, this documentation arrives during a particularly volatile period. Google's AI features have been prioritising NHS content for health queries, government sources for regulatory information, and established news outlets for current events—creating new visibility challenges for commercial sites that previously ranked well for informational searches.

The Five Core Sections of Google's Guide: What They Actually Mean

Section 1: Is SEO Still Relevant for Generative AI Search?

Google's answer is unequivocal: yes. The guide states that "our generative AI features on Google Search are rooted in our core Search ranking and quality systems." This matters because it contradicts the theory—popular in early 2024—that AI search would require completely separate optimisation strategies focused on "speaking to the LLM" rather than ranking in traditional results.

What this means practically: if your content ranks in the top 10 for relevant queries, you're already in the pool of sources Google's AI can cite. The guide confirms that AI features use "retrieval-augmented generation and query fan-out to highlight content from our Search index." Translation: the AI retrieves pages that would rank well for the query and related queries (that's the "fan-out"), then generates an answer based on those retrieved pages.

In our experience at HeroSEO, we've observed strong correlation between traditional ranking position and AI citation likelihood. Pages ranking in positions 1–5 for queries that trigger AI Overviews appear as cited sources significantly more often than pages ranking lower. The pattern is clear: improving your traditional rankings through solid SEO fundamentals—relevance, authority, user satisfaction signals—remains the highest-leverage activity for AI search visibility. This reinforces what Google's guide emphasises: there's no separate "AI SEO" discipline that differs from quality SEO practice.

Section 2: Apply Foundational SEO Best Practices

This section reiterates principles from Google's existing quality guidelines: create helpful, reliable content for people; ensure your site is crawlable and indexable; use structured data appropriately. Google's emphasis here is deliberate—they're signalling that the fundamentals haven't changed, even if the presentation layer has.

The guide specifically mentions E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) as remaining central to AI search visibility. Pages demonstrating clear expertise and authoritativeness appear to be cited more frequently in AI Overviews, particularly for YMYL (Your Money or Your Life) topics like health, finance, and legal information.

What the guide doesn't emphasise enough: structured data has become more valuable, not less. While Google states you shouldn't create LLM-specific markup, existing schema types—especially FAQ, HowTo, Product, Review, and LocalBusiness—appear to increase citation likelihood in AI features. We've tested pages with properly implemented FAQ schema against similar pages without it, and the correlation with AI citation frequency is substantial.

Section 3: Mythbusting—What You Don't Need to Do

This is the most valuable section for practitioners, because Google explicitly rules out tactics that had gained traction in the industry. The guide states you should not:

• Create content specifically optimised for LLM comprehension (no "talking to the AI")
• Use special markup or tags to make content more "AI-friendly"
• Attempt to "prompt engineer" your content by including phrases like "as an expert" or "comprehensive guide"
• Build separate pages targeting AI search versus traditional search
• Add hidden text or metadata intended for AI systems to read

Why does Google discourage these approaches? Because they don't work with how the system actually functions. Google's AI features aren't reading your content the way ChatGPT reads a prompt. They're using traditional information retrieval to find relevant, high-quality pages, then using language models to summarise and synthesise that information. The retrieval step uses the same signals as organic search ranking—relevance, authority, freshness, user satisfaction metrics.

Section 4: Explore Agentic Experiences

This forward-looking section introduces "agentic" AI—systems that can take actions on behalf of users, not just provide information. Google describes scenarios where AI Mode might help a user book a restaurant reservation, schedule a service appointment, or complete a purchase without leaving the search interface.

For UK businesses, this has significant implications. You'll need to ensure your booking systems, APIs, and transaction platforms can interface with Google's agentic systems. The guide doesn't provide technical specifications yet (those are coming in a separate developer documentation update), but it signals that structured data for Actions, Reservations, and Orders will become increasingly important.

Insurance brokers, solicitors, accountants, and other service businesses should start preparing now. If your quote request or consultation booking process requires multiple page loads, form submissions, or manual email exchanges, you're at risk of being bypassed when AI agents can complete bookings directly with competitors who've implemented streamlined, API-accessible systems.

Section 5: Next Steps—What to Focus On

Google's final section offers general recommendations: monitor your Search Console data, stay updated on changes, focus on user satisfaction. It's the weakest section of the guide, because Google doesn't yet provide AI-specific metrics in Search Console. You can't see which of your pages are being cited in AI Overviews, how often, or what queries trigger those citations.

This is where third-party tools become essential. We use a combination of AI visibility tracking platforms and manual SERP monitoring to understand our clients' AI search performance. The methodology we've developed (detailed in the audit framework section below) provides the actionable data Google's guide conspicuously lacks.

What Google's Guide Deliberately Omits: The Critical Gaps

Google's documentation is valuable for what it confirms, but more revealing for what it avoids addressing. These omissions aren't oversights—they're strategic choices to maintain flexibility or avoid committing to metrics that might change as AI features evolve.

No Measurement Framework

The guide doesn't explain how to measure your AI search visibility, track citation rates, or benchmark performance against competitors. Google Search Console doesn't currently distinguish between clicks from AI Overviews versus traditional results. You can't see impression data for AI features. Third-party tools like Semrush, BrightEdge, and specialised AI visibility platforms have attempted to fill this gap, but coverage is inconsistent and UK-specific data remains limited.

No Citation Preference Factors

While Google confirms that AI features use core ranking systems, the guide doesn't explain why some pages ranking in similar positions get cited whilst others don't. In our analysis, we've identified several factors that appear to increase citation likelihood beyond ranking position: content freshness (pages updated within 90 days show stronger citation patterns), presence of author bylines with credentials, specific use of FAQ and HowTo schema, and domain authority in the specific topic area. Google's guide doesn't acknowledge any of these patterns.

No UK-Specific Guidance

The guide is written for a global audience and doesn't address regional variations in AI search behaviour. UK users receive AI Overviews less frequently than US users for many query types. NHS and GOV.UK content dominates health and regulatory AI responses even when commercial pages rank higher in traditional results. GDPR considerations affect what data can be used in agentic experiences. None of this appears in Google's documentation.

No Competitive Displacement Discussion

Google avoids addressing the elephant in the room: AI Overviews reduce click-through rates to source websites. Publishers and content sites have reported traffic declines for queries that now trigger AI features. Google's guide focuses on visibility within AI features but doesn't discuss strategies for maintaining overall traffic when featured snippets, AI Overviews, and other SERP features increasingly answer user queries without requiring clicks.

Our approach at HeroSEO with clients has been to focus on building authority in topics where AI can't fully answer the question—complex, multi-step processes; personalised advice requiring interaction; tools and calculators; and bottom-funnel commercial queries where users still want to compare detailed options. These query types either don't trigger AI Overviews or generate high click-through rates when they do.

The UK Perspective: How AI Search Behaves Differently in Britain

Google's guide presents AI search optimisation as a universal practice, but UK businesses face distinct challenges and opportunities that American-focused guidance doesn't address. Understanding these regional differences is essential for developing an effective strategy.

Government and NHS Content Prioritisation

For any health-related query, AI Overviews in the UK heavily favour NHS content, even when commercial health providers rank well in traditional results. We've tracked health queries over time and found NHS sources heavily represented in AI Overviews. This isn't bias—it's Google applying appropriate E-E-A-T weighting to authoritative public health sources.

Similarly, regulatory, legal, and compliance queries prioritise GOV.UK content. Commercial solicitors and compliance consultants need to position their content as complementary rather than competitive—explaining how regulations apply to specific situations, providing implementation guides, or offering tools that help users apply government guidance to their circumstances.

British English and Cultural Context

AI language models are trained predominantly on American English content, which occasionally produces awkward or incorrect terminology in UK AI Overviews. This creates an opportunity: content that uses British English consistently and addresses UK-specific contexts has a higher likelihood of being cited for UK searchers.

Ensure your content uses British terminology (redundancy not layoffs, holiday not vacation, solicitor not lawyer, accountant not CPA) and addresses UK-specific frameworks (HMRC not IRS, FCA regulations, UK GDPR, Companies House requirements). This isn't just about spelling—it's about semantic accuracy that helps Google's retrieval systems identify your content as the most relevant for UK queries.

Local Business AI Visibility

AI Mode shows particular promise for local service queries in the UK. When users search for "plumber near me" or "emergency electrician Manchester", AI Mode can integrate Maps data, reviews, availability information, and service descriptions in ways traditional local pack results can't. But this requires proper LocalBusiness schema implementation, consistent NAP (Name, Address, Phone) data across directories, and crucially, direct booking or quote request capabilities that AI agents can access.

UK local businesses should prioritise schema markup for opening hours (including bank holidays), service areas (using UK postcodes and city names, not American ZIP codes), accepted payment methods, and emergency service availability. These structured signals dramatically increase the likelihood of appearing in AI-generated local recommendations.

The Comprehensive AI Search Optimisation Audit Framework

Google's guide tells you what to focus on but not how to assess your current state or prioritise improvements. This framework provides a systematic approach to auditing your site's readiness for AI search visibility, organised by priority level.

Priority 1: Foundation Audit (Complete This First)

Priority 2: Content Quality Assessment

Review your top-performing content (by traffic and rankings) against these criteria. Each page should score at least 4 out of 6 to be considered AI-ready:

• Comprehensive coverage: Does the page answer the primary query plus 3–5 common follow-up questions?
• Clear authorship: Is there an author byline with credentials or expertise demonstrated?
• Current information: Has the page been updated within the last 6 months? Are dates clearly displayed?
• Cited sources: For factual claims, are sources linked and authoritative?
• Structured formatting: Does the content use semantic HTML headings, lists, and tables appropriately?
• Scannable layout: Can a user (or AI) quickly extract key information without reading every word?

Pages scoring 0–2 should be rewritten or consolidated. Pages scoring 3 should be enhanced with additional signals (author credentials, FAQ schema, citations). Pages scoring 5–6 are strong candidates for AI visibility—monitor them specifically.

Priority 3: Schema Implementation Checklist

Based on our testing across 40+ UK client websites, these schema types have the strongest correlation with AI Overview citations:

1. FAQ Schema: Implement on any page that naturally addresses multiple related questions. Use actual questions users ask (check "People also ask" boxes and Answer the Public for UK-specific phrasings).
2. HowTo Schema: Essential for process-oriented content. Each step should be specific and actionable. Include images where possible (HowTo schema with images shows stronger patterns in our data).
3. Article Schema: Include author, publisher, datePublished, dateModified. Use actual Person schema for authors with credentials, not generic Organisation attribution.
4. Product Schema: For e-commerce and service pages. Include offers, reviews, and availability data. For services, use aggregateRating if you have genuine reviews.
5. LocalBusiness Schema: Critical for multi-location businesses. Include opening hours with UK bank holiday exceptions, service areas using UK postcodes, and accepted payment methods.
6. Review Schema: When you have authentic reviews. Never use fake or incentivised reviews—Google's quality systems detect patterns and will demote the entire domain.

For UK businesses in regulated sectors (financial services, legal, healthcare), include Organisation schema with regulatory credentials (FCA number, Solicitors Regulation Authority ID, CQC registration). These signals increase E-E-A-T assessment and improve AI citation likelihood for regulated topics.

Priority 4: AI Visibility Monitoring

Since Google doesn't provide AI-specific metrics, build your own monitoring system:

Manual SERP tracking: Identify your 20–30 most valuable queries. Check them weekly in incognito mode from UK IP addresses. Document when AI Overviews appear and whether your content is cited. Track the pattern—is AI Overview presence increasing? Are your citations increasing or decreasing?

Traffic pattern analysis: In Google Analytics or your analytics platform, segment queries that trigger AI Overviews (you'll need to identify these manually first) and track whether their click-through rates decline over time. If you're maintaining or growing clicks despite AI Overview presence, your content is compelling enough to drive clicks even when an answer is provided. That's the goal.

Competitor citation tracking: For your core topics, note which competitors get cited in AI Overviews when you don't. Analyse what those pages do differently—longer content? Better structured data? More authoritative author credentials? More recent updates? Use competitive insights to improve your own pages.

Share of voice estimation: If you're cited in 40% of AI Overviews that appear for your target queries, and your top competitor is cited in 60%, you've identified a visibility gap worth addressing. Prioritise enhancing content for queries where competitors are out-citing you.

Preparing for Agentic AI: What UK Businesses Need to Do Now

Google's guide introduces agentic experiences—AI taking actions on users' behalf—but provides minimal implementation guidance. Based on Google's public statements at developer conferences and our analysis of early agentic search patterns, here's what UK businesses should prioritise now to be ready when these features expand.

Booking and Transaction Systems

If your business takes bookings, appointments, or reservations, ensure your system supports:

• Real-time availability checking via API (not just form submissions that require human response)
• Automated confirmation without manual approval steps
• Calendar integration (Google Calendar, Outlook) for seamless scheduling
• Structured reservation schema on your booking pages
• Clear cancellation policies marked up with schema

Professional services firms—solicitors, accountants, consultants—should implement proper appointment booking systems rather than relying on "contact us for a consultation" forms. When AI agents can book directly with a competitor but must send an email to you, the friction disadvantage is significant.

Product Information and Inventory

For e-commerce businesses and retailers, agentic AI will require:

• Real-time inventory data accessible without authentication
• Detailed product specifications in both human-readable and schema formats
• Dynamic pricing with clear availability dates
• Delivery options and costs calculable via postcode without requiring cart creation
• Return policies and warranty information clearly structured

UK businesses should ensure pricing includes VAT in displayed amounts (unlike US practice of showing pre-tax prices) and that delivery information references Royal Mail, DPD, and other UK carriers rather than American services.

Service Quotes and Lead Generation

For B2B services and complex products requiring quotes, the challenge is greater. Agentic AI can't complete a sale that requires custom pricing, but it can pre-qualify leads and gather information. Consider implementing:

• Quote calculators that provide genuine estimates (not "contact us" barriers)
• Interactive needs assessments that help users self-identify requirements
• Tiered pricing frameworks that give realistic ranges even if final pricing requires conversation
• Chatbot qualification that can answer 80% of preliminary questions without human involvement

The principle is reducing friction. If an AI agent helping a user can get a meaningful quote range from your competitor's calculator but gets a "we'll contact you" message from your site, you've likely lost that prospect.

Legal and Compliance Considerations

UK businesses must consider GDPR implications of agentic AI. If Google's AI agent books an appointment on behalf of a user, who is the data controller? What consent is required? How is data handled if the booking is cancelled? Google hasn't provided detailed guidance on these questions yet, but UK businesses should:

• Ensure your privacy policy explicitly covers automated bookings and third-party agents acting on user behalf
• Implement clear consent flows that work even when booking happens outside your website
• Maintain audit trails showing what information was provided at booking and what consent was given
• Ensure your systems can handle subject access requests that include AI-mediated transactions

Working with a GDPR compliance specialist to review your booking and transaction flows before implementing agentic-ready systems is advisable, particularly for businesses in financial services, healthcare, or other regulated sectors.

Adapting Your Content Strategy: From Keywords to Topics to Answers

Google's guide confirms that AI search represents an evolution in how content is discovered and presented, which requires rethinking traditional content strategies. The shift isn't from "keywords" to "no keywords"—it's from individual keywords to topical authority, and from singular pages to interconnected content ecosystems.

The Topic Cluster Model for AI Visibility

AI Overviews tend to cite multiple pages from the same domain when that domain demonstrates comprehensive topical coverage. Rather than creating isolated pages targeting individual keywords, structure your content as interconnected clusters around core topics relevant to your business.

Example for an insurance broker: instead of standalone pages for "business insurance quote", "professional indemnity insurance", and "public liability insurance", create a comprehensive business insurance pillar page with detailed sub-pages for each insurance type, all interlinked and using consistent terminology. Add supporting content answering common questions: "What's the difference between professional indemnity and public liability?", "How much business insurance do I need?", "Does my business insurance cover working from home?"

This cluster approach mirrors how AI features work—they're more likely to cite multiple pages from your site if you've demonstrated authority across related concepts. We've seen improved visibility for businesses with mature topic clusters compared to those with isolated keyword-targeted pages.

Answer-First Content Structure

Traditional SEO content often follows an inverted pyramid: introduction, background context, then eventually the answer. AI-optimised content should answer the question in the first 2–3 sentences, then provide supporting detail, alternative perspectives, and related information.

This doesn't mean making content shallow. It means frontloading the core answer, then going deeper for users (and AI systems) that want comprehensive understanding. Structure your content like this:

1. Direct answer (50–75 words): Answer the title question clearly and completely
2. Key points (100–150 words): Expand on the answer with 3–5 essential points
3. Detailed explanation (500–800 words): Provide depth, examples, and context
4. Related questions (200–400 words per question): Address 3–5 common follow-up questions
5. Practical application (300–500 words): Explain how to use this information
6. Expert perspective (100–200 words): Add nuance, caveats, or industry-specific insights

This structure serves both AI citation (which often pulls from the direct answer and key points sections) and user experience (users who want depth can continue reading).

Update Frequency and Content Freshness

Google's guide doesn't emphasise this, but we've found strong correlation between recent content updates and AI citation likelihood. Pages updated within 90 days show stronger patterns than pages with older publication dates, even when traditional rankings are similar.

Implement a systematic content refresh programme: review your top 50 pages quarterly, update statistics and examples, add new FAQs based on recent customer questions, and ensure all dates are clearly displayed. Don't just change the "last updated" date—make substantive improvements that genuinely enhance the content's current relevance.

For UK businesses, be particularly attentive to regulatory and legislative changes. When the FCA updates guidance, HMRC changes tax thresholds, or new employment law takes effect, update relevant content immediately. Being the first to reflect current information in your content dramatically increases AI citation likelihood for queries related to that change.

Measuring Success Without Google's Data: Creating Your Own AI Visibility Metrics

The most frustrating aspect of Google's guide is its silence on measurement. Search Console doesn't differentiate AI Overview clicks from organic clicks. You can't see which queries trigger AI features for your pages or how often you're cited versus competitors. Until Google provides these metrics (if they ever do), you need to build your own measurement framework.

The Hybrid Measurement Approach

Combine manual tracking, indirect signals, and third-party tools to estimate your AI visibility:

Citation frequency tracking: Manually check your 30–50 most valuable queries monthly. Record whether AI Overviews appear and whether your content is cited. Calculate a simple "citation rate"—if AI Overviews appeared 40 times and cited your content 18 times, your citation rate is 45%. Track this over time to identify trends.

Share of voice estimation: For queries where AI Overviews appear, document all cited sources. If your content appears alongside 3 competitors in an AI Overview, you have 25% share of voice for that query. Aggregate across all tracked queries to estimate overall AI share of voice in your topic area.

Traffic pattern analysis: Segment your analytics data by query categories. Compare click-through rates for queries that consistently trigger AI Overviews versus those that don't. If CTR declines for AI-heavy queries but your citation rate is strong, it indicates AI Overviews are providing sufficient answers to reduce clicks—that's expected. If CTR declines but you're rarely cited, you have an optimisation problem.

Brand mention tracking: Use tools like Brand24, Mention, or custom Google Alerts to track when your brand is mentioned in online discussions about your industry topics. AI language models are trained on web content—including forums, social media, and comments. Building brand association with your core topics through genuine engagement increases the likelihood of future AI mentions.

Leading Indicators of AI Visibility

Because direct AI metrics are difficult to track, focus on leading indicators that correlate with eventual AI visibility:

• Traditional ranking improvements: As confirmed by Google's guide, better traditional rankings lead to higher AI citation likelihood. Track ranking movement for your core queries.
• Featured snippet wins: Pages that earn traditional featured snippets show higher propensity for AI citations. Monitor your featured snippet presence in Search Console.
• Schema validation: Regularly check that your structured data remains error-free. Schema errors correlate with reduced AI visibility.
• Content depth metrics: Track average word count, number of FAQ sections, use of data and citations. Pages with these elements consistently outperform shorter, less comprehensive content in AI citations.
• Topical authority scores: Use tools like Semrush's Topic Authority metric or build your own by tracking your ranking presence across all queries in your core topic areas. Higher topical authority correlates with more frequent AI citations.

Competitive Benchmarking

Assess your AI visibility relative to competitors using this framework: