Two things are true at once in a lot of businesses right now, and together they are getting perfectly good marketing cancelled.
Search traffic is falling. Buyer intent is rising.
In HubSpot's 2026 State of Marketing report, 58% of marketers said AI referral traffic carries much higher intent, with the visitors who do arrive already further along in their buying journey. The same survey found nearly 30% of marketers reported decreased search traffic as people turn to AI tools. Fewer visitors, warmer people.
Judge that on a traffic chart and it looks like decline. Judge it on revenue per visitor and it can look like the best year you have had.
Most businesses are looking at the traffic chart.
Falling search traffic is not bad news by itself. If enquiries, sales and revenue per visitor are holding or rising, you have lost the browsers and kept the buyers. Change what you measure and who you write for before you panic about the graph.
Where the missing searches went
They moved upstream, into a conversation you cannot see.
The old journey had five or six searches in it. What is this, how does it work, what does it cost, who does it near me, is this company any good, and finally the branded search that arrived at your door.
Now the first four can happen inside an assistant. The buyer describes their situation in a paragraph, gets a comparison, asks two follow ups and forms a view. Then they search once, for the business they have already half chosen.
So you lost four searches and kept the one that was worth having.
Google is part of this too. When the results page answers the question, the click never happens. We cover how that plays out in Google just ate your traffic. Either way the research still gets done, somewhere your analytics has no view of.
What you lost was the browsing. The buying is still there, and one buyer costs the same to serve as one browser did.
Why this breaks your reporting
Every standard marketing report is built on volume. Sessions, impressions, clicks, followers, reach. All of it assumes that more attention at the top produces more sales at the bottom, at a stable rate.
The rate is no longer stable.
Which produces a specific and painful failure. Traffic halves, enquiries hold steady, revenue goes up, and the marketing budget gets cut anyway because the graph everyone looks at in the meeting is pointing down.
I have watched businesses cancel the channel that was quietly closing their best work because the dashboard made it look sick.
Volume numbers were always a proxy. They were a decent one while the conversion rate held still. Now a volume report is closer to a vanity metric than a business metric, the trap we pull apart in vanity metrics are quietly bankrupting your marketing.
The numbers to run instead
Four metrics, and you can pull every one of them from tools you already have.
- Revenue per session. Start here. If sessions fell 40% and revenue per session doubled, you are winning and the traffic chart is lying.
- Enquiry to sale rate. When buyers arrive further along, this rises. A jump here is the fingerprint of demand that was educated before it reached you.
- Branded search volume. When an assistant recommends you, a search for your name is often where it lands. Search Console now has a branded queries filter for this. Check it monthly.
- Sales cycle length. Better informed buyers decide faster. If the gap from enquiry to close is shrinking, the upstream work is doing its job even though no report shows it.
How to set these up in an afternoon
- Revenue per session: export monthly sessions from your analytics and monthly revenue from your accounts. Divide one by the other in a spreadsheet. If you sell offline, use enquiries per session as a stand-in.
- Enquiry to sale rate: count enquiries and closed sales per month from your CRM or inbox. Even a rough count works.
- Branded search: in the Search Console performance report, filter to branded queries (or to queries containing your business name) and record impressions and clicks each month.
- Sales cycle: for each closed sale, note the date of first contact and the date of purchase. Average the gap monthly.
- Add one question: ask every new enquiry how they found you, with "an AI assistant like ChatGPT" as a named option.
That last one matters more each month. Visits from AI apps, as opposed to AI websites, get categorised as direct traffic in GA4, so the only reliable way to catch them is to ask. There is more on tracking that traffic in traffic from AI converts better and almost nobody is tracking it.
What to change about the work itself
Stop producing top of funnel explainers. The "what is" and "how does it work" content that used to bring the volume is now answered before anyone arrives. Writing more of it is supply for a demand that moved.
Write for the person who already understands. Comparisons. Selection criteria. What goes wrong. Real pricing. When you are the wrong choice. A definition can be summarised away. A judgement call from someone who does the work cannot.
Make your sales process match the buyer. If prospects arrive having done four rounds of research, a discovery call that opens with "so tell me a bit about your business" wastes the advantage. Skip ahead. Ask what they have already ruled out and why.
Raise your standard for what counts as a lead. With fewer, better prospects, every hour spent chasing an unqualified one hurts more. Tighten the form. Ask a qualifying question.
A worked example: the accounting firm with the falling graph
Say you run a small accounting practice in Canberra. Your monthly report shows website sessions sliding steadily over the past year. Your blog posts explaining "what is a BAS" and "how does GST work" used to pull in most of your traffic, and they have dropped hardest. Your business partner wants to cut the SEO budget.
Before anyone decides, you run the four numbers. Imagine enquiries have stayed roughly flat, the share of enquiries that become clients has gone up, more new clients mention they "asked ChatGPT for a local accountant", and branded searches for the firm's name have crept up. Revenue per session has climbed because the visitors who remain are the ones ready to engage an accountant.
The decision flips. You keep the budget, stop writing beginner explainers and rewrite your three key service pages for business owners comparing firms: who you are best for, what a typical engagement includes, your fee ranges and what changes them, and when a different kind of firm would suit them better. You change the first call to start with "what have you already looked at?" This is a hypothetical, but it shows how the same data can justify opposite decisions depending on which line you look at.
The strategic read
If your business has real expertise, this shift is good news.
The old game rewarded whoever could produce the most content and buy the most clicks. Volume beat quality with depressing reliability, because volume caught people early and nurtured them into buyers.
The new game takes much of that early stage out of your hands. What is left is the part where a human decides who to trust with something that matters. They decide on evidence, on how specific you are and on your reputation, which is ground a good small business can win on.
A content mill has nothing to offer a buyer who has already read forty summaries and now wants to know who has done this before.
The pricing conversation nobody expects
One more consequence catches businesses off guard.
Buyers who arrive after four rounds of AI research often arrive knowing the price range in your category. They have been told what things typically cost, sometimes accurately and sometimes not.
That undercuts the old approach of holding price back until you have built value in a meeting. The value has to be built before the conversation, in material a machine can read.
Publish your pricing structure, or at least your ranges and what drives them. It filters out the wrong prospects and gives assistants accurate information to repeat. The people who do reach you have already seen the number.
If you are nervous about publishing prices, start with what makes a job cost more or less. That answers a lot of what a buyer wants to know, and it makes you the honest option in a category where competitors hide behind "contact us for a quote".
What to do this month
Pull twelve months of data. Chart sessions and revenue on the same page. Then chart revenue per session on its own.
If that third line is going up while the first is going down, print it and take it to whoever is worried about the traffic chart.
Then take the three pages that generate your enquiries and rewrite them for someone who already knows what they want and is deciding between you and two others.
If you want help rebuilding the reporting and the pages, that is exactly what our digital marketing and AI search team does at LIQID NOISE. We set up the numbers that show what is really happening, then write for the buyer who has already done the research.
Frequently asked questions
Why is my website traffic dropping but my sales are not?
Many early-stage questions are now answered by AI assistants and search result summaries before anyone clicks. The people who still visit tend to be further along the buying journey, so fewer visits can produce the same or more sales. Check revenue per session and your enquiry to sale rate to see whether the traffic you kept is simply worth more.
What should I measure instead of website traffic?
Track revenue per session, enquiry to sale rate, branded search volume and sales cycle length. Together they show whether visitors are arriving better informed and more ready to buy. Add a "how did you find us?" question to every enquiry, including AI assistants as an option, because visits from AI apps can show up as direct traffic in analytics.
Should I still write blog posts if search volume is falling?
Yes, but change what you write. Basic "what is" explainers are increasingly answered before anyone reaches your site. Write for buyers who already understand the topic: comparisons, selection criteria, real pricing, common failures and honest guidance on when you are not the right fit. Those pages need judgement, which a summary cannot replace.
Should small businesses publish their prices online?
In most cases, publishing price ranges and what drives them helps. Buyers often arrive already knowing typical costs from AI research, so hidden pricing feels evasive. Published ranges filter out poor-fit prospects, give AI assistants accurate figures to repeat, and mean the people who contact you already know roughly what you charge.