The traffic story is the wrong story

Picture the meeting. Third Tuesday of the quarter, the marketing update, twelve minutes on the agenda. You have the Looker Studio dashboard up, the one that pulls from GA4 and Search Console. Organic sessions are down again, third quarter running. Semrush says the site still ranks where it ranked. And a partner who has never once asked about the website leans in and says: "So is this AI thing hurting us or helping us?"

You do not have a clean answer, because the question is built on a metric that stopped meaning what it used to mean, and none of the tools on that dashboard were built to notice.

The story everyone is selling

Every pitch you have received this year tells the same story. AI search is the new traffic source. Get cited and the visitors follow.

It is a comfortable story because it fits the stack you already have. GA4 for sessions, Search Console for impressions, Semrush or Ahrefs for the ranking check, HubSpot for the leads that came out the other end. If AI is a traffic channel, you measure it the way you measure everything else, and the report writes itself.

The trouble is that the story is wrong, and research published last month says so.

What the researchers found

Last month, two universities, Penn and Northeastern, published a study on how people use Google now. They recruited 1,100 people and, for a week, watched what those people clicked.

They set it up like a test. One group got Google with the AI answers switched off, links only. One group got Google's AI Mode for every search, the version that answers the question instead of showing links. The rest got Google as it is today.

When people got the AI answer, they clicked through to websites far less, about 19 fewer clicks for every hundred searches. Take the AI away and clicks went up, about 9 more per hundred. People with the AI answer also said they trusted what they found less.

That is the whole finding, and it is a simple one. The more Google answers the question itself, the fewer people leave Google. Your website is on the other side of that click.

So what is the click for

Your website was where a client came to check you out. The credentials, the practices, the lawyers, the matters you have handled. The visit was the vetting.

That vetting is now happening inside the answer, and in more ways than one. A GC who got your name from a trusted referral checks it in private, four questions deep, before anyone calls you. A new CFO at a client you have served for a decade types "is this a reputable firm" before approving next year's budget. The night before your engagement renews, a board member asks "any red flags with this firm." Or a client describes a problem, a letter of credit that will not be honoured, and the answer names firms on the way to solving it. Someone asks who leads the specialty, and the partner who built your book is not named. And yes, a prospect asks who handles cross-border insolvency in New York, and gets a list of three or four names.

Your next client asks AI who to hire. Your current clients ask AI whether to keep you. In none of these does anyone need to visit your site. The answer did the checking.

Put the study next to those scenes. It says that when AI answers the question, people stop clicking through. So the visits you lost did not vanish. The checking moved somewhere your dashboard cannot see. Search Console will still count the impressions and call it a good month. It cannot tell you whether you were named. Measured by traffic, AI looks like it does not matter. Measured by whether you are on the list, it is the only thing that matters.

The dashboard that answers the partner

If the click is not the metric, three questions are. Put them on one page and take that to the next meeting.

Are we named? On the questions above, the ones your clients actually ask about your practices and your partners, not a generic prompt list. A firm can be named for one practice and invisible for the one that pays the bills; one firm-wide score hides that. Measure the firm, the practice, and the lawyer separately, because that is how the engines answer.

In whose words? In a 1,600-answer Perplexity study, a third of the firms AI named had no page of their own among the sources. They were described in other people's words, sometimes wrong ones. Retired partners still listed. Offices that do not exist. Fixing the record is faster than winning a shortlist, and a partner can verify every fix.

Does it hold? In that same study, roughly half the firms named appeared in only one of three identical runs. A single check tells you about a Tuesday. Benchmarks frozen on day one and re-measured monthly tell you whether you are there.

You can get a rough first version of that page this afternoon, without anyone's help. Open ChatGPT and Google's AI Mode. Type the four or five questions your own clients would ask: who handles your lead practice in your city, whether your firm is reputable, who leads the specialty your best partner built, any red flags. Read who comes back. If your firm is missing from an answer where a rival is named, you have your first line for the partner meeting. What you will not get this way is the part that holds up in a room: the same questions on every engine, run more than once, against the firms the partners actually name, and re-run next month so movement is real.

Back to the meeting

The partner asks whether AI is hurting or helping. You say: it is hurting us in a place this chart cannot show. Clients used to come to our website to check us out. Now they ask ChatGPT or Google and get three or four names. If we are not one of them, we are not considered, and nothing in this report would tell us it happened. The question is whether we are on that list when a client asks, and today I cannot tell you.

Bring that page next time, in place of the traffic chart. Which of our practices AI names when a client asks. Which firms it names instead of us. Whether that changed since last month.

That is a marketing update a partner can do something with. The old one was a weather report.

I should say where I sit. We built Selectio because the version of that page that holds up in a room did not exist. Nothing in the stack above can tell you whether your firm, a practice, or a partner is named, so we built the thing that does, and measuring it is the whole job. If you want to see what the page looks like for your firm, before anyone talks about a plan, ask and we will run it.

---