Google's AI search optimization guide: what brands must do now

By Stephan Charles | Last fact-checked: <time datetime="2026-06-14">June 14, 2026</time>

On <time datetime="2026-05-15">May 15, 2026</time>, Google published its first official guide for optimizing content to appear in AI Overviews and AI Mode, under a new "Generative AI fundamentals" section in Search Central. A <time datetime="2026-06-05">June 5</time> update named GEO and AEO as legitimate SEO services. At the same time, Google extended its spam policies to cover AI citation manipulation, adding penalties for brands that game AI answers. These three moves give brands both a framework and a warning: earn citations using Google's published approach, or face lower rankings or removal from results. BrandCited tracks citation rates across 9 AI engines so brands can measure where they stand before those consequences arrive.

BrandCited's free scan checks your citation rate across 9 AI engines in 30 seconds, with each gap ranked by impact. Run it at brandcited.ai.

What did Google publish about AI search in May and June 2026?#

Google published "Optimizing your website for generative AI features on Google Search" on May 15, 2026. It's the first time Google released official guidance for appearing in AI-generated answers, filed under the "Generative AI fundamentals" section in Search Central documentation, alongside guidance on Core Web Vitals and mobile-first indexing.

On June 5, 2026, Google updated its long-standing "Do you need an SEO?" hiring guide to add GEO and AEO as named, legitimate service categories. Search Engine Journal's coverage of the June 5 update confirmed the framing: from Google's perspective, "optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

Google's generative AI features use retrieval-augmented generation to pull from the standard search index, which means AI citation visibility and organic ranking share the same technical foundation. Google's guide also confirmed that llms.txt, content chunking, and AI-specific rewriting are not required for its AI features — what works for organic search works for AI citations.

Why do search rankings no longer predict AI citations?#

Rankings and AI citations now measure different things. The data from early 2026 makes this concrete.

In mid-2025, 76% of AI Overview citations came from pages ranking in Google's organic top 10. By early 2026, that number fell to 38%, according to ALM Corp's analysis of the citation pool shift. A page ranking 11th can appear in an AI Overview when it uses the right schema and content structure. A page ranking 1st can be cut when its content doesn't match AI retrieval patterns.

AI Mode compounds the gap further. Authoritytech.io's citation analysis found that AI Mode and AI Overviews share only 13.7% of cited URLs. A brand's AI Mode visibility and its AI Overview visibility are nearly independent metrics, even though both live inside Google. Superlines' analysis of 34,234 AI responses puts the domain-level overlap at 21.9%.

The practical consequence: a brand's search rank tells you almost nothing about its AI citation rate. Both metrics need separate tracking and separate optimization. BrandCited monitors AI citations across 9 engines as a metric independent of organic rank.

What do Google's spam policies now say about AI citation manipulation?#

Google extended its spam policies to AI Overviews and AI Mode on May 15, 2026. The updated policy covers "attempting to manipulate generative AI responses in Google Search." Search Engine Land's report on the update confirmed that the same penalties apply as for traditional organic spam: lower rankings or removal from search results.

The practice Google named is "recommendation poisoning," which covers biased "best-of" listicles written to push AI models toward a specific brand, and paid placements presented as organic AI citations. Google's enforcement applies the same algorithms as traditional spam detection.

Brands that have purchased AI citation placements or commissioned biased comparison content now carry enforcement risk. A brand doesn't need to have manipulated results to be investigated — the content pattern itself triggers the review.

Run a free BrandCited scan to see where your citations are coming from across 9 engines: brandcited.ai

How do Preferred Sources in AI Mode change brand citation visibility?#

Google's Preferred Sources feature expanded to AI Overviews and AI Mode on May 27, 2026. When a user marks a publisher as preferred, Google cites that publisher more often in AI answers on that user's account, with a "Preferred" badge appearing next to the citation. PPC Land's coverage of the rollout confirmed that Google treats preferred status as an active input into citation selection, not a display label.

SE Roundtable's analysis of the feature reported it now operates across AI Overviews, AI Mode, and Top Stories globally, in all languages where those features run.

For brands, Preferred Sources creates a new metric: how many users have marked you as preferred across their Google accounts. That number doesn't appear in Search Console yet, but its effect shows up in citation rates over time. Brands with loyal readers who use Google's preference tool will see their AI citation frequency climb on those users' devices, separate from any change in organic rankings or content strategy.

The "Preferred" badge is visible only to the user who set the preference, and Google confirmed the signal influences citation frequency rather than serving as a display indicator alone.

What should brands do now to earn AI citations?#

These steps are ranked by citation impact, highest first.

1. 1Run a BrandCited scan to get your baseline citation rates across 9 AI engines. You need the current numbers before you can set a target. The scan takes 30 seconds: brandcited.ai
2. 2Add statistics with named sources to your top 10 pages. Superlines found that content with statistics achieves 30–40% higher AI citation rates than comparable pages without them. The source link matters as much as the number itself.
3. 3Refresh your three most-cited pages. Pages updated within the last two months earn 28% more citations than older content, according to Superlines. Updating means new data, corrected statistics, and a refreshed dateModified in Article schema.
4. 4Add FAQPage schema to your homepage and top product pages. Write each FAQ as a question a user would type into AI search. Here's the JSON-LD format Google reads:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What does [brand] do?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Your complete answer here, 2-4 sentences. No external references required."
    }
  }]
}

1. 1Remove paid citations and biased comparison content. Google's spam policy update targets recommendation poisoning. Content created to push AI models toward your brand rather than to inform users now carries enforcement risk under the May 2026 update.
2. 2Skip llms.txt for Google AI features. Google's official guide says it's not required. Spend that time on FAQPage schema, content freshness, and named author authority instead.

How does BrandCited audit AI citation readiness?#

BrandCited's audit engine checks AI citation readiness across 30+ checks covering entity schema, content structure, author authority, and citation frequency. The audit returns a score from 0 to 100 for each of 9 AI engines, with each failing check ranked by impact. BrandCited's checks map to Google's published AI optimization guide, so brands can see which of Google's own recommendations they're currently missing.

Run a free BrandCited scan to see your citation rate across 9 AI engines before Google's enforcement rules affect your results. The scan takes 30 seconds and shows every gap ranked by impact: brandcited.ai

AI search news from the past 24 hours#

• Anthropic overtakes OpenAI in US enterprise adoption: The Ramp AI Index shows Anthropic surpassed OpenAI in enterprise software spend share among US companies for the first time, as of June 2026. (Ramp AI Index)
• Gemini 3.5 Flash as default in AI Mode globally: Google confirmed Gemini 3.5 Flash, announced at Google I/O on May 19, is now the global default model in AI Mode. (TechCrunch)
• OpenAI files confidential S-1: OpenAI submitted its S-1 registration to the SEC on June 8, 2026, opening the process toward a potential IPO. (OpenAI)
• AI referral traffic conversion rate: AI-sourced sessions convert at 14.2% versus 2.8% for traditional organic search, according to the Opollo 2026 AI Search Benchmark Report.
• Google spam enforcement confirmed: Google confirmed on June 13 that its spam policies now apply to brands attempting to manipulate AI Overviews and AI Mode, with enforcement active. (Search Engine Land)

Frequently asked questions#

Google's May 15, 2026 guide says brands should focus on traditional SEO fundamentals: clear content structure, accurate entity information, and FAQPage schema for question-and-answer content. The guide confirms that llms.txt, content chunking, and AI-specific rewriting are not required for Google's AI features. Pages crawlable by standard search bots don't need separate AI crawl directives.

On June 5, 2026, Google updated its vendor hiring guide to add GEO and AEO as named, legitimate service categories. Google's framing: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." For brands evaluating AI optimization vendors, this gives them a Google-endorsed standard for assessing vendor claims.

Yes. Google updated its spam policies on May 15, 2026 to cover "attempting to manipulate generative AI responses in Google Search," including biased comparison content and paid placements presented as organic AI recommendations. Penalties include lower rankings or removal from results, the same consequences that apply to traditional organic spam.

Preferred Sources is a Google Search feature that lets users mark publishers as preferred. When a user does this, that publisher gets cited more often in AI Overviews and AI Mode for that user's account, with a "Preferred" badge appearing next to the citation. Google expanded the feature to AI Overviews and AI Mode on May 27, 2026, making it the first user-controlled signal that directly inputs into AI citation behavior.

AI Overviews and AI Mode weight schema markup, content structure, named author authority, and content freshness over raw link authority. A page ranking 11th can earn AI citations with FAQPage schema and a named author byline. A page ranking 1st gets cut if it lacks those signals. ALM Corp documented the drop from 76% to 38% at the URL level between mid-2025 and early 2026. AI Mode and AI Overviews share only 13.7% of cited URLs even within Google's own products, according to Authoritytech.io.

Run a free BrandCited scan at brandcited.ai. The scan checks citation rates across 9 AI engines in 30 seconds and returns every gap ranked by impact, mapped to the specific audit check that explains the gap. No signup required to start.