Cited Is Not Clicked: What an 87% Drop in AI Click-Through Means for Your Brand

A court brief filed on September 17, 2026 put a hard number on a soft fear. Links shown inside AI chat get clicked far less than links in search.

The brief comes from news publishers suing OpenAI and Microsoft. It quotes a 2023 message from an OpenAI engineer: “no matter how prominently we show the links, users won’t click.”

The numbers

The brief cites Microsoft data. It compares click-through rate for the same sites in Bing Chat and in Bing web search.

Publisher group	Click-through rate in chat vs. search
The New York Times sites	87% to 93% lower
Daily News sites	83% to 91% lower
Ziff Davis sites	51% to 94% lower

Read these numbers with care. They come from one side of a lawsuit. The court has not ruled. The brief does not say when the data was collected or how many queries it covered. OpenAI and Microsoft dispute the case.

The direction is still clear. The best case in the table is a 51% drop. The worst is 94%.

Citations and clicks are two metrics

The tools that measure AI search count citations. Bing Webmaster Tools reports citation share. Microsoft says that number is not traffic share. Google Search Console’s AI report shows impressions. It does not show clicks.

So a brand can be cited often and visited rarely. Both are true at once.

This changes what a citation is worth. In search, a ranking was a door. The user walked through it to your page. Your page told the story. In AI chat, the answer tells the story. The user reads it and moves on. Your page is a source, not a destination.

Google is moving the same way

On September 28, Google opened AI Mode’s monitoring feature to all users worldwide. A user tells AI Mode what to watch. Search keeps checking sites, forums, social posts, and a Shopping Graph of more than 60 billion products. It sends an update when something changes.

Every example Google gave was local or retail. New restaurants. Pop-up markets. Price drops. Back-in-stock alerts.

Google has not said if these updates carry links. It has not said how it picks sources. A local shop cannot yet tell how often an update will send a visit.

A user who gets a price alert may never open your site. They act on what the alert says. If the alert is wrong, they act on a wrong fact.

What to measure instead

If the click is going away, measure the answer.

• Accuracy. Does the answer state your price, hours, and features correctly?
• Completeness. Does it name the products you sell today? Does it leave out your newest one?
• Consistency. Do ChatGPT, Gemini, and Perplexity say the same thing about you? Or three different things?

A wrong answer is worse than a wrong page. With click rates 51% to 94% lower, few users will click through to find the fix.

Fix the sources the engines read

Engines build answers from pages they can parse. Start with your own site.

• Put key facts in plain text near the top of the page.
• Add structured data for prices, hours, and locations.
• Label every form field and button. On the top one million home pages, 51% have form inputs with no label. An AI agent cannot use a field it cannot name.
• Update third-party listings when your facts change.

Then ask the engines what they say. Write down every error. Repeat each month.

The click was the old proof that search worked. The answer is the new proof. beket.ai checks that answer across ChatGPT, Gemini, and Perplexity, scores it against your real data, and fixes what it can. See what AI says about your business at beket.ai.