Measuring SEO as zero-click grows: 4 metrics tied to revenue

This article is for you if you pay for SEO every month and are starting to see reports that don't match reality: the ranking table is still green, but fewer calls are coming in. We won't get into optimization techniques here, only measurement. The four SEO metrics below are the ones we use for this very site and for the SEO projects we take on for clients, along with how to pull the numbers using free tools and a few places where the numbers are easy to misread.

Where old-style SEO measurement breaks: rankings no longer tell you about revenue

Ranking was always a proxy metric. Nobody cares whether they're in position three or five; they care about how many customers call. Ranking got used because it used to track clicks fairly closely, and clicks tracked customer numbers. That chain has just broken in the middle.

Ahrefs reported on April 17, 2025 that the click-through rate of the number one position dropped 34.5% for queries with an AI summary. Pew Research Center tracked the real browsing activity of 900 American adults and found the click rate on a result was only 8% when the page had a summary, compared with 15% when it didn't. Clicks on links inside the summary itself were just 1%.

Read together, those two numbers say something simple: keeping your position no longer means keeping your customers. Old-style SEO measurement used ranking as the main yardstick, and now it measures the wrong thing. We cover the full context of this phenomenon in What is Zero-Click Search.

Impressions up and clicks down is the new normal

Open Google Search Console for any content website from the last two years and you'll often see a scissors-shaped chart: the impressions line climbing while the clicks line stays flat or slopes down. Many SEO agencies look at that chart, report the part going up and skip the part going down.

That scissors shape is the result of machines answering right at the top of the page, not a technical fault on your website. Your content is still being served, and may even be read and used by the machine; readers just no longer need to click. So a report that only shows off impressions hides the part that matters.

One point of fairness, which we're stating here so nobody can use it against you: as of August 2026, Search Console folds AI answer data into the Performance report and doesn't break it out as its own line for comparison. That means there's no way to read off directly “how many of my impressions were inside an AI summary.” The four metrics below exist to work around that gap.

Three metrics to stop putting at the top of your report

Keep these three for reference when you need them, just don't let them be the headline metrics anymore.

• Average ranking across your keyword set. A single number blended from hundreds of queries with very different behavior now means even less. Keep it for technical monitoring, but don't use it to score performance.
• Total organic traffic. This number is falling in almost every industry, even for websites doing better than last year. Score yourself on it and every year looks like a failure.
• Number of posts per month. This is an activity metric, not a results metric. Publishing 20 posts when none of them makes it into the machine's answers is worth zero.

Metric 1: how often your brand is named in the machine's answers

This metric directly replaces ranking, and you can measure it by hand without buying any tools. The method is to build a fixed set of questions and rerun exactly that set once a month, counting how many times your company name appears.

The question set should have about 30 to 40 questions in three groups. Group one is buying questions with a location, like “reputable website design company in Ho Chi Minh City.” Group two is comparison questions, like “which warehouse management software should a distribution company choose.” Group three is expert questions in your own trade. Keep this question set the same from month to month; change the questions and you lose the ability to compare.

Run the question set in three places to cover the field: ChatGPT, Google's AI Mode and Perplexity. For each question, record three things in a spreadsheet: whether your name is mentioned, in what position in the list, and which other names appear alongside yours. That last column is your real list of competitors in the machine's eyes, and it's often very different from the competitor list your sales team has in mind.

A morning is enough to finish 40 questions the first time. From the second month on it's much faster, about an hour, because the sheet is already set up. We have one person on the team do this on the 5th of every month, as fixed as closing the books.

The number to track is your presence rate: how many questions, out of the total, mention your name. The first month is usually shockingly low, and that's the real number. To raise it, the work lies in credibility signals, which we cover separately in E-E-A-T in the AI era.

Metric 2: how many people search for your company name directly

You can pull this metric for free from Google Search Console, and it's the most honest yardstick for the value that clicks no longer reflect. A reader sees your name in the machine's answer, doesn't click right away, but a few days later types your company name directly to find you again. That search falls into the branded query group.

Here's the short way to get the number. In Search Console, open the Performance report, set a query filter containing your brand name, choose the last 3 months and compare with the same period before. Remember to include misspelled and no-diacritic variants, since customers often mistype proper names. Record two numbers: total impressions and total clicks for the branded group.

Read the numbers by direction, not by single month. If branded queries rise steadily for three months in a row while total organic traffic stays flat, that's a good sign: people are getting to know you somewhere else and then coming back to type your name. On the other hand, if the branded group stays flat for half a year, then no matter how good your rankings are, your brand still hasn't gotten into anyone's head.

A common trap: a company runs a big ad campaign and branded queries spike, then drop back when the ads stop. Note the start and end dates of ad campaigns right on your tracking sheet, so next month you don't celebrate by mistake.

Metric 3: visits from AI assistants, separated from regular search

Readers do click the links inside the machine's answers, just not many. Those visits reach your website as referral sources, and if you don't separate them, they get lost among other sources and vanish from your report.

In Google Analytics 4, set up a custom channel group and put these source domains into one bucket: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai. Name that bucket “AI assistants” and let it stand on its own, at the same level as organic search and social. It takes about 15 minutes to set up, and you do it once and use it for good.

The volume in this bucket will be small. For most Vietnamese business websites today it usually sits at around 1 to 3% of total visits. Seeing that number, many people want to drop it altogether. We advise against that. Look at metric 4 before you decide.

What's worth tracking in this bucket is the month-over-month growth rate, not the absolute number. Going from 0.4% to 1.2% in six months is a very clear direction, even though the number is still small. And what's worth doing right away is checking which pages those visits land on, because those are exactly the pages the machine is citing, which makes them the model for the next pages you write.

Metric 4: value per session, instead of session count

This is where the way you read the numbers flips, and where we say the opposite of many agencies. Less traffic isn't automatically bad. Readers have already had the basics answered by the machine, so those who still click are past the first round; they come to verify or to buy. Visitor quality has changed, so the yardstick has to change.

There are three numbers to pull, all available in GA4 if conversion tracking is set up. One is the conversion rate per session from organic search. Two is revenue or number of contact forms per hundred sessions. Three is the conversion rate of the AI assistants bucket from metric 3, separated from the rest.

The third number often surprises people. The AI assistants bucket makes up only 1 to 3% of visits, but its conversion rate is usually much higher than regular search, because someone who clicks from an answer has already read the introduction before clicking. Once you can measure this, the budget meeting gets much easier: a small bucket, but a valuable one worth feeding.

A tidy way to present it is a four-row table, one row per source, with three columns: sessions, conversion rate and value per hundred sessions. One glance shows which source is feeding the company, and it's also the table to argue over when reallocating your SEO budget next quarter. We set up conversion tracking properly as part of SEO services, and if your website can't yet attach conversion events, the groundwork has to be fixed first, see website design service.

Pull your SEO metrics into a one-page report for your boss

A 12-page SEO measurement report won't get read. We compress it to exactly one page with four blocks, each with one big number and an arrow comparing it to the previous period.

• Block 1. Presence rate in the machine's answers, in the form “12 out of 40 questions,” with the three competitors named most often alongside you.
• Block 2. Impressions and clicks of the branded query group, compared with the same 3-month period before.
• Block 3. Sessions from the AI assistants bucket, with its percentage of the total and the three most-visited pages.
• Block 4. Value per hundred sessions, split by source.

Read blocks 1 and 3 monthly, and blocks 2 and 4 quarterly. The reason is that the last two move slowly, and reading them monthly invites seasonal noise and hasty decisions.

One piece of advice we're willing to be wrong about: don't wait until you have three months of data to start measuring. Take your baseline this month, even knowing it doesn't look good, because the most valuable thing about this set of metrics is the trend line, not the number from any single month. Starting three months late means losing three data points for good. If you're deciding whether to keep funding an in-house SEO team, also read the SEO profession in 2026 before you decide.

SEO measurement by industry: which metrics matter most for each trade

These four metrics apply to every industry, but the weights differ, so each trade's SEO scorecard looks different.

• High-value services such as construction, industrial equipment and software. Metrics 1 and 4 matter most, because customers research for a long time before calling, and each order is large enough that one quality session outweighs hundreds of casual ones.
• Retail and e-commerce. Metrics 3 and 4 matter most, measured by product group. Supporting platform: ecommerce website design.
• Restaurants, spas and location-based services. Metric 1 matters most, with questions that include district names, plus the number of calls from your business profile on the map. Manage it with restaurant management software or spa management software.
• Education and training. Metric 2 matters most, because parents and students almost always look up the center's name again before paying. Platform: AI-integrated LMS from Mona.Software.
• Manufacturing and distribution. Metric 1 matters most, with technical comparison questions, because buyers often ask the machine before drawing up a supplier shortlist. Paired with warehouse management software.

To push the number in metric 1 up, the work lies in how you write: answers must be short and sit right at the top of a section so the machine can excerpt them. We walk through that writing method step by step in writing answer-first content, and the part about declaring information for machines to read is in what is Schema.