AI News Β· Edition 49 Β·
The visit that never happens
Bing's own numbers suggest AI answers are swallowing website clicks. How to check whether it is happening to you, without buying anything.
The one that matters
AI answers are quietly cutting website visits β check your own numbers before believing anyone
The advertising trade publication PPC Land reports that internal Bing data shows Microsoft's Copilot assistant cut click-through rates to publishers by as much as 94%. In plain words: when somebody asks an assistant a question instead of typing it into a search box, the answer appears in the chat window, and the visit to the website that supplied that answer often never happens.
Start with what that number is and is not. It is one figure, from one company's internal measurement, reported by one trade outlet. "Up to 94%" is a worst case, not an average. And "publishers" means news and content sites β businesses whose entire model is selling attention on a page. A physiotherapy clinic is not a publisher. Do not treat 94% as your number. Treat it as a direction of travel that you can measure yourself.
What it actually means for your business
The question that matters is narrower than "is search dying". It is: how many of your enquiries begin with somebody searching? Consider three real shapes of business:
- A 600-student school where admissions enquiries spike each January and February, and a good share of first contacts come from parents searching for schools in the area. If the assistant answers "which schools near me take Class 6 admissions" by summarising three school sites, two of those schools get described without being visited.
- A 30-bed clinic ranking for "physiotherapy near me" or "dengue test timings". The intent does not disappear β it gets satisfied inside the chat, using whatever facts the model can find about your timings and fees.
- A 40-staff workshop whose PDF parts catalogue quietly brings in trade enquiries when buyers search a part number. A model that cannot read that PDF cleanly will quote whichever competitor published the same spec as plain text.
The pattern is the same in all three: fewer visits, same intent, and the answer assembled from sources you do not control. Your website stops being a destination and becomes a supply source that something else reads on your behalf.
What it does not mean, and who is overstating it
It does not mean your website matters less. It means it is read by machines more and humans less β which is close to the opposite of "shut it down". It also does not justify a redesign; a beautiful site whose fees, timings and address live inside an image or a PDF is less useful now than an ugly one with those facts in plain text.
The overselling has already started. Agencies are packaging "AI search optimisation" or "GEO" retainers at roughly βΉ40,000ββΉ80,000 a month in India, or CA$1,500β3,000 in Canada, promising placement inside ChatGPT or Copilot answers. Nobody can guarantee that. There is no auction to bid in and no ranking dashboard to buy. Anyone promising a position is selling you a guess. The opposite crowd β "nothing has changed, keep buying links" β is also wrong, just more cheaply.
What a sensible owner does this month
- Measure before reacting. Open Google Analytics and Search Console, compare the last 12 months of organic visits against the same period a year earlier, and β separately β count actual enquiries. Traffic down with enquiries flat is a different problem from enquiries down. Cost: one afternoon, zero rupees.
- Make your facts readable. Fees, timings, address, services, phone number, admission dates β as ordinary text on ordinary pages. Not only in a brochure PDF, a poster image, or an Instagram post.
- Keep your Google Business Profile correct. For local businesses it is still the most-cited source about you.
- Reduce the dependence. A current WhatsApp or email list of parents, patients or buyers is a channel nobody can re-rank. Build it while the search traffic still exists.
If your enquiries do not start with a search, none of this is urgent. Find that out first.
Also worth knowing
How this briefing is put together
Every morning we read the dayβs AI announcements and reporting from the companies themselves and from the technology press, then pick the handful that actually change something for a working business. The analysis is ours and it is written for owners and managers, not engineers. Every story links to its original source β read them, and disagree with us where weβve got it wrong.