What Google just published about AI search, and what it means for your firm

A general counsel sits down on a Tuesday morning with a hard matter: a cross-border IP dispute for a technology client, on a tight timeline. She needs outside counsel.

In 2022, she would have called three peers for referrals. In 2026, she opens ChatGPT and types: "Which firms handle cross-border IP litigation for mid-market technology companies?"

Thirty seconds later, she has a shortlist. Five firms, each with citations. She is already opening their websites.

Your firm might be on that list. Or it might have been considered and dropped. You would not know which. On May 15, 2026, Google published a guide that describes how that list gets built.

---

What Google just told us about how the shortlist gets built

Google's document is titled Optimizing your website for generative AI features on Google Search. It is the company's most direct statement so far on how a firm earns a place in AI answers. Two ideas in it should change how firms think about their marketing, and one line the document is careful to soften deserves a closer look.

Idea one: one question becomes many

Google now formally describes what it calls query fan-out: when you ask its AI one question, it runs several related searches at once, then assembles a single answer from across them. The person asking sees only the finished answer.

Apply that to a legal question. A general counsel asks, "Best M&A firm in New York for a cross-border tech deal." Behind the scenes, the AI runs several smaller searches in parallel: top M&A firms in New York, firms with cross-border deal experience, firms serving mid-market technology companies, recent tech M&A rankings, partner-level deal history.

Each of those searches pulls its own sources. A directory names a few firms. A legal news outlet names a few more. Deal write-ups and partner credentials round it out. The AI then looks for the firms that show up consistently across all of them, with the strongest supporting evidence. A couple make the cut. The rest get dropped. The general counsel sees one shortlist and has no idea several searches just ran.

This is the central shift. For twenty years, search was about ranking for the query. Now your firm needs to show up, credibly and consistently, across the smaller questions the AI asks on the way to the answer. A firm that ranks far down on the headline term but shows up strongly across those smaller questions can be the firm that gets named.

Whether your firm appears on the shortlist is not a measure of quality. It is a measure of how consistently you show up across the questions AI is asking.

Jacob Shamis, Co-Founder, Selectio.ai

Idea two: what Google says you can stop worrying about

The guide also includes a myth-busting section. Google tells site owners they do not need to:

• Create llms.txt files, AI text files, or special AI markup.
• Break content into small pieces for AI to consume.
• Rewrite content specifically for AI systems.
• Pursue inauthentic mentions across the web.
• Overfocus on structured data.

If a vendor is pitching any of these as essential AI optimization services, Google has now said in writing that they are not. The real work, building a consistent and trusted footprint across the sources AI reads, is harder and slower than markup tricks.

---

What Google has good reason to downplay

One line in the guide deserves close attention. Google addresses the term AEO, Answer Engine Optimization, directly and concludes that, from its own perspective, optimizing for generative AI search is "still SEO."

That holds inside Google's own framing. Google is talking about its own AI surfaces, AI Overviews and AI Mode. For those, the practices that have always driven Google performance still apply.

There is also a commercial reality behind that framing. Search advertising is the largest part of Google's business. Every minute a general counsel spends inside ChatGPT, Perplexity, Gemini, or Claude is a minute she is not spending on Google. Telling the market "this is all still SEO" keeps the world organized around Google's view of search.

Five AI engines matter for whether your firm gets recommended: ChatGPT, Perplexity, Gemini, Claude, and Google's AI Overviews. Each gathers and weighs its sources differently. Google's guide covers Google Search. Measuring and improving your visibility across all five is the work.

Three things this means for your firm

For the partner or marketing lead trying to act on this, the guide and the changes behind it come down to three shifts.

• Depth beats keywords. The old playbook treated content as a series of bets on individual keywords. The new one treats your practice pages, attorney bios, case results, and articles as a single body of evidence the AI reads together.
• Your authority is spread across many places: directories, legal news coverage, deal write-ups, bar profiles, speaking engagements, and your own insights page. The AI connects these. If your footprint is thin or inconsistent across them, excellent legal work will not translate into AI visibility.
• Your dashboards will not warn you. Google Search Console folds AI traffic into the standard report, and ChatGPT, Perplexity, Claude, and Gemini do not appear there at all. Your firm could be invisible across four of five engines and your monthly dashboard would look fine. The tools were built for a different era of search.

And it runs on its own

Four days after the guide, at Google I/O on May 19, 2026, Google's CEO described the redesigned Search as the biggest change in more than 25 years. The new search box, with AI Mode behind it, is rolling out wherever AI Mode is available. When a general counsel searches Google now, she increasingly sees a synthesized answer with firm recommendations built in, produced by the same one-question-becomes-many mechanic.

Google also previewed information agents that keep working after the search is over: a partner can ask one to follow a litigation matter, a regulatory change, or a deal market, and it returns to the web on its own to report what is new. Firms that publish citable work steadily will appear in those reports. Firms that publish thin or sporadic work will not, because the agent has nothing to cite when it comes back. An invisible firm no longer loses one search at a time. It gets filtered out on a loop.

---

What to do first

Three steps, in order:

• Map where you show up. Before fixing anything, learn whether your firm appears when AI runs its smaller searches: your city, your practice area, your matter type, your recent results. These are different searches pulling from different sources, and firms often perform unevenly across them without knowing where the gaps are.
• Audit your footprint. Your directory profiles, your legal news coverage, your attorney profiles, your firm bio pages. AI reads these together and looks for them to agree. Where they are thin, outdated, or describe your core practice areas inconsistently, they leave gaps in the coverage that gets you named.
• Start before it feels urgent. The standing AI uses to build its shortlists accumulates, and firms that start earlier hold the lead.

The first step is measurement. You cannot fix what you cannot see.

---