On May 15, 2026, Google published its first official guide to optimizing websites for AI Overviews and AI Mode. The most useful section is everything it tells site owners not to do.

Four “AI optimization” tactics that have been driving agency up-charges for the past two years are now, per Google itself, optional at best:

• llms.txt files and other “special” markup. Google’s crawler may discover them, but they get no preferential treatment in the index.
• Content “chunking.” Google says its systems can extract the relevant passage from a multi-topic page on their own — pre-fragmenting articles for AI is not necessary.
• Rewriting content specifically for AI. AI features already understand synonyms and meaning. Long-tail keyword variation rewrites are not necessary.
• Overfocusing on schema for AI purposes specifically. Schema still drives rich results in traditional Search; Google is not telling you to remove it. It’s telling you to stop pretending it’s an AI-specific lever.

The plain-English translation for managing partners: if your “GEO agency” is billing you premium hours for llms.txt setup, content chunking, or AI-specific content rewrites pitched as the secret to AI Overviews, you are paying for theatre.

All that said, “not required” is not the same as “worthless,” and the real snake oil isn’t llms.txt itself; it’s agencies billing premium retainer hours for tactics that should be one-time, low-effort, set-it-and-forget-it deliverables. The 9Sail position: stop paying agency rates for cosmetic AI tactics. Keep paying for the foundations Google explicitly validated.

Good SEO Was Always Good GEO

This guide is not a repudiation of AI optimization work. It’s a repudiation of the wrong AI optimization work. Everything Google tells you to keep doing maps cleanly to the foundations that have driven organic visibility for the last decade.

Google’s affirmative list:

• unique non-commodity content with first-hand experience,
• clear technical structure,
• crawlability and page experience,
• accurate local business and entity details, and, where it fits the business
• early preparation for agentic experiences like browser agents and the Universal Commerce Protocol.

That list maps directly to the foundations we laid out in our 5 Pillars of AI and Generative Engine Optimization: authority architecture, structured content, citation-worthy originality, multi-platform entity recognition, and technical infrastructure.

The cleanest framing of the guide is also the most quotable one: good SEO has always been good GEO. The platforms generating AI answers, including Google’s own AI Overviews, pull from the same Search index your SEO work has been building all along. Google has now said this on the record.

For your firm’s budget, that means two things. If you’ve been investing in traditional SEO foundations like technical health, attorney-attributed long-form content, digital PR, and directory authority, that work is producing more value than ever, because it’s now feeding both blue-link rankings and AI citations from the same index. The dollars you’ve been spending on tactical “AI hacks” sold as separate from SEO are sunk costs. Stop the bleed; redirect the budget.

The “non-commodity content” line Google drew is the same line that legal buyers draw

Google’s guide explicitly contrasts commodity content—Google’s own example is “7 Tips for First-Time Homebuyers”—with non-commodity content, like a first-person case story from someone who has actually done the thing.

Apply that to legal. A generic “What is a non-compete?” article is commodity content. A partner’s analysis of how the FTC’s non-compete rule changed the negotiation posture in their last three executive separations is not. A 1,200-word restatement of black-letter law on Jarkesy is commodity. A litigator’s view on how Jarkesy is reshaping their firm’s SEC enforcement defense calendar is not.

This is the same content that earns CLO and GC trust on a referral check. Sophisticated buyers are not impressed by black-letter restatement, and neither is the model deciding whether to cite your firm to them.

The Two Nuances Law Firms Specifically Cannot Ignore

Google’s Guide is About Google Search, Not the Entire AI Ecosystem

The most important caveat in this entire conversation: Google’s guide is explicitly about Google Search’s AI features. It is not a global statement about how every AI platform behaves.

When a general counsel asks Perplexity, ChatGPT, or Gemini to help them validate outside counsel, those systems are indexing and weighting signals differently than Google does. Entity authority, citation footprint across legal directories, and structured firm data still influence those answers even when Google says you don’t strictly need them for AI Overviews specifically.

Attorney, LegalService, and LocalBusiness schema still help in three places: traditional rich results in Google Search, cleaner entity recognition across non-Google AI platforms, and disambiguation when multiple firms share similar names. Reading Google’s guide as permission to strip schema is a misread of what it actually says.

“Not required today” is not the same as “won’t matter tomorrow”

The compounding effect on indexed assets is the part the lazy take misses. Schema markup, structured author bios, and entity data don’t just produce immediate value, they accumulate authority every time they’re crawled and re-indexed. A firm that waits for Google to officially endorse a signal before deploying it is, by definition, behind every firm that deployed it two years ago.

This is the same logic that explained why early SEO investors won so decisively in 2010–2015. The cost of being early is small. The cost of being late is enormous because you’re competing against a competitor’s compounded index footprint.

Apply that lens to the four tactics in Google’s guide and the picture clarifies fast:

• Schema markup already produces rich-result value, the indexing benefit compounds, and the likelihood it grows in importance is high. Keep investing.
• llms.txt has no value today, but deployment takes minutes and the file may become a citation signal as the AI ecosystem matures. Set it and move on. This is not a billable engagement.
• Content chunking and AI-specific rewrites have negative ROI today, no compounding benefit, and no credible path to becoming required. Skip.

The same logic applies to content structure. You don’t need AI-specific chunking. You do still want answer-first paragraphs, clear H2/H3 hierarchy, and well-organized FAQ-style sections, not because they’re a SERP feature (Google retired FAQ rich results on May 7, 2026), but because human readers benefit from skimmable structure, and AI systems lean on the same structural signals that help humans skim.

Where your AI optimization dollars should actually go in the next 12 months

First-hand expertise content attributed to individual attorneys. Google’s “unique point of view” criterion lines up with how AI tools surface named attorneys before firm names in recommendation queries. Byline your work, and make the byline a real attorney with a real bio page, real credentials, and real publication history.

Entity consistency across the legal directory ecosystem. Chambers, Legal 500, Best Lawyers, Martindale, Justia — these are the high-credibility sources non-Google AI platforms cross-reference when they’re deciding whether to cite your firm. Inconsistent NAP data, mismatched bios, and stale practice descriptions across these sources weaken the model’s confidence in your firm as a citable entity.

Digital PR in practice-area-relevant publications. Earned mentions in niche industry trades are what AI models read as topic-specific authority signals. A WSJ mention helps brand awareness; an ABA Practice of Law publication mention helps citation likelihood.

Technical health and Core Web Vitals. Google reiterated that crawlability and page experience remain foundational for generative AI eligibility. A site that won’t render quickly on mobile cannot get cited, regardless of how good its content is.

Measurement infrastructure. Track AI citation frequency across ChatGPT, Perplexity, and Google AI Overviews on a consistent cadence. You cannot reallocate budget intelligently if you cannot see where you currently appear, and “AI visibility” as a single blended number is a vanity metric. Break it out by platform.

How to Pressure-Test Your Current AI Optimization Spend This Quarter

Five questions to take into your next conversation with any current or prospective AI optimization vendor:

1. “How much of our current scope is dedicated to llms.txt, AI-specific rewrites, or content chunking?” After Google’s May 2026 guide, what is the case for keeping any of it as billable retainer work?
2. “How are you measuring AI citation frequency across ChatGPT, Perplexity, and Google AI Overviews — separately, not as one blended number?” If they can’t show you platform-level visibility data, they don’t have a measurement system.
3. “What percentage of our content investment is attributed to individual attorneys versus the firm institutionally?” If the answer is “we publish under the firm byline,” they’re not building the entity authority that AI systems reward.
4. “What is your plan for legal directory entity optimization — Chambers, Legal 500, Best Lawyers — as AI citation sources, not just rankings submissions?” If directories are still being treated as a once-a-year submission task, you’re missing where non-Google AI is sourcing its answers.
5. “Can you connect AI visibility movement to actual business development outcomes — new matters, referral validation, branded search lift — within six to twelve months?” If the deliverable is “we got you cited,” but no one can connect it to the matter pipeline, you’re buying impressions.

A vendor that treats AI optimization as a separate discipline from SEO is, after May 15, 2026, selling you redundancy. If you want the longer breakdown of how the two disciplines actually integrate, the SEO vs. GEO framework walks through it.

If you want to see specifically where your current AI optimization spend is and isn’t producing citations, 9Sail’s AI Visibility Audit benchmarks your firm against peers and shows the citation data underneath the impression numbers.

The Bottom Line for Law Firm Marketing Leaders

Google didn’t kill AI optimization. It killed the lazy version of AI optimization that a lot of agencies have been billing for since ChatGPT launched.

The work that actually moves the needle is the same work 9Sail has been pushing in our 10 things to know about GEO and our 5 Pillars framework all along.

If your current vendor’s roadmap leans heavy on the four tactics Google just officially deprecated, the next conversation to have isn’t with them. It’s with someone who’ll show you the citation data and reallocate the budget.

Request an AI Visibility Audit. Your move.