Your Firm Can Be The Top AI Recommendation and Invisible at The Same Time

Almost everyone in my industry talks about AI visibility like it’s a single point – the next generation of “ranking at the top of Google” is just one simple line item of “get recommended by AI.” But unlike Google Search, “AI” is not a single-source monopoly.

That’s what makes AI recommendations such an incredible opportunity for law firms. ChatGPT, Claude, Google AI Mode, Gemini, Copilot, Perplexity, and Grok represent multiple platforms that are potential lead sources. These platforms are all different.

We are running experiments and gathering data through our AI marketing platform, Sequoia, all the time. As mentioned before, the vast majority of that intel is kept secret and pushed into our platform to benefit our clients. As I continue to fight a war against BS vanity metrics, we’ve shared some of our intel in the “Knowledge is Law Firm Marketing Power” feature of our free resources section.

We have never publicly shared any of our AI data until now. When running a research hypothesis, we saw something in the crosstabs that was just too interesting to keep to ourselves.

The three most popular AI platforms can’t agree on which lawyer to recommend.

We Analyzed 107,716 After Asking for Law Firm Recommendations in the Top 100 U.S. Cities

We took 100 of the largest cities in the country and 32 practice areas, everything from car accident and criminal defense to probate and wrongful termination, and built 3,200 questions phrased the way an actual person looking for an actual lawyer would phrase them. Then we ran all 3,200 through ChatGPT, through Claude, and through Google AI Mode. That is 9,600 questions and 107,716 citations after cleanup. The full study is published here with the complete methodology if you want to see the full analysis.

Across all of it, those three platforms recommended 19,692 different law firms.

They agreed on 578 of them – less than 3% of the time.

ChatGPT, Claude, and Google AI Mode Agree on the Right Lawyer Just 2.9% of the Time

That is the number that stopped me. Of every law firm recommended anywhere in this study, 2.9 percent were recommended by all three platforms. Google AI Mode alone named 9,778 firms that neither ChatGPT nor Claude mentioned a single time across 3,200 questions.

Narrow it to one question, where all three are answering the identical prompt about the identical city, and the overlap between any two of them lands between 2.8 percent and 8.6 percent. Claude and Google AI Mode returned nothing in common at all in 70.6 percent of questions.

Keep in mind that this is a single-point-in-time snapshot. These systems can recommend somewhat different answers to the same question at different times which we’ve seen in our massive dataset from our AI Monitoring platform. But variance does not carry you from a third down to under three percent, and it does not explain the next part at all.

These Three Platforms Have Different Favorite Resources

The three platforms are not drawing from some shared pool of legal directory listings. Each one has its own favorites, and the favorites barely touch.

• Avvo appears in 87.1 percent of Claude’s answers. On Google AI Mode it appears in 5.4 percent.
• Martindale shows up in more than half of everything Claude says and in roughly one out of a thousand Google AI Mode answers.
• Super Lawyers runs the exact opposite direction, present in 62.8 percent of Google AI Mode answers and completely absent from Claude.
• Google AI Mode also pulls from Yelp and Reddit on a regular basis. Neither chat platform touched either one in our sample.

So when a firm asks which directories matter for AI, the question is missing a word. Which directories matter for which platform. A thin Avvo profile is a Claude problem. An unclaimed Super Lawyers listing is a Google problem. They are separate projects with separate payoffs, and keeping those profiles current only earns you anything on the platforms that actually read them.

Some Practice Areas Are Wide Open Right Now

The disagreement is not spread evenly. Corporate law produced the most agreement at 45.2 percent, car accident at 32 percent, personal injury just under 30. Those are areas where the platforms clearly share some sense of who the real firms are, which also means the field is more settled.

Now the other end. Estate planning produced zero. Across 100 cities and three platforms, the systems never once converged on a firm they all agreed to name. Probate came in at 2.5 percent and trust work at 3.3 percent.

My read is that transactional practice areas have not built the layer of third party validation that injury and criminal work spent twenty years building, so the systems have nothing solid to anchor to and end up pulling from three unrelated pools. But either way the practical fact holds. If you do estate planning work, nobody has won this yet, and the door is open wider than anything I have seen in organic search in a decade.

What Law Firms Can Learn

Stop reporting AI visibility as one number. There is no single position to win, so there is no single number that means anything. A firm checking ChatGPT and liking what it sees has learned something true about ChatGPT and close to nothing about anywhere else. Any tracking worth paying for covers every platform your potential clients might reasonably open.

Be skeptical of anyone claiming to be the top AI recommendation, including yourself. On our data that claim is true on one platform and false on the others about nine times out of ten. Usually, the person saying it believes it, because they checked one place and stopped. This is exactly the vanity metric problem wearing new clothes, and it will send your budget somewhere useless.

Move now, because this will not stay this open. Of the firms in our study, 15,308 appeared in exactly one city. There is no established national name these systems default to, which is genuinely unusual and will not last. Right now, consistent and verifiable information about your firm in the places each platform actually looks is still enough to get you named.

Seven platforms are recommending lawyers by name. We measured three of them and found almost no agreement, which means seven separate opportunities rather than one crowded one. That is the whole argument for treating answer engine optimization as real work instead of a line item, and it is the reason we finally shared an AI discovery with the public.

Hopefully, we can get ahead of the false information spreading about how law firms are actually represented in AI search.