SEO Metrics & KPIs: What to Measure Now That the Old Numbers Are Broken

Does this look familiar?

Organic traffic is down. Impressions are volatile. Rankings swing around for no obvious reason. Your keyword lists keep shrinking.

And yet… your enquiries, demos and sales are mostly fine.

SEO teams everywhere are being told performance is collapsing, while the business doesn’t see a collapse at all.

I’ve given a talk about exactly this, and the core of it fits in two sentences: the problem isn’t your SEO. It’s your metrics.

In this post, I’ll break down why the classic SEO metrics stopped making sense, which KPIs still matter, and what I now measure instead.

What are SEO metrics and KPIs?

Quick definitions first, because the two get used interchangeably and they shouldn’t be.

SEO metrics are anything you can measure: impressions, clicks, rankings, referring domains, engagement, conversions.

SEO KPIs are the handful of metrics you’ve deliberately chosen because they connect to what the business actually cares about.

Every KPI is a metric, but very few metrics deserve to be KPIs. Most bad SEO reporting comes down to ignoring that difference - filling dashboards with everything measurable and calling it insight.

Why your SEO numbers don’t make sense anymore

The old SEO measurement stack rested on three assumptions:

• Impressions = visibility
• Search volume = demand
• Rankings = opportunity

That stack only worked because we had stable access to Google’s SERPs. We don’t anymore - and each assumption has broken in its own way.

Impressions dropped without visibility dropping

In late 2024, Google effectively removed the ability to request 100 results at once, and started requiring JavaScript to access search properly in early 2025.

Reason being, Google wanted to shut out bots and scrapers - but a side effect was that huge amounts of non-human impressions disappeared from Search Console overnight.

Here’s the detail most people missed: Search Console counts an impression when a URL is loaded in the results, not when a human actually sees it. Rank-tracking tools loading page after page of results were inflating impression counts for years.

So many sites saw impressions fall off a cliff while their page-1 presence didn’t change at all. If your impressions chart nosedived around then, it may well have been your data getting cleaner, not your visibility getting worse.

Search volume was never a fact

We treated search volume like a fact for years. It was always a model - inferred from SERP accessibility, clickstream panels and extrapolation.

When Google restricted deep results and enforced JavaScript, the inputs those models relied on disappeared. The tools didn’t suddenly get worse; their raw material did.

Rank tracking is genuinely harder now

Two rank-tracking tools can check the same keyword on the same day and disagree - and both can be “right.”

SERPs now vary by user, location, session and layout test. AI Overviews appear for one query today and not tomorrow. You’re not ranking on a stable page anymore; you’re appearing inside an interface that reshapes itself constantly.

As I put it in the talk: rankings didn’t die. Our ability to measure them did.

The most important SEO performance metrics and KPIs

None of this means measurement is pointless - it means you have to be much more deliberate about what you treat as a KPI.

The way I see it, your KPIs should sit as close to money as possible:

• For B2B and service businesses: organic enquiries, calls, demos and the pipeline they create. Track form submissions and calls from organic, and put a value on them.
• For ecommerce: organic revenue and its share of total revenue - not sessions.
• For everyone: non-branded organic clicks (your growth engine), branded search demand (your brand equity), and conversion rate from organic traffic.

Watch out for SEO vanity metrics

A vanity metric is any number that looks impressive in a report but doesn’t connect to a business outcome.

They’re seductive because they usually go up - and a chart that goes up feels like progress, even when nothing about your leads or revenue has changed.

The classic SEO vanity metrics include:

• Organic traffic without context - more visitors means nothing if they’re the wrong visitors
• Impression totals - especially unreliable now, for all the reasons above
• Keyword counts - “we rank for 5,000 keywords” says nothing about whether any of them matter
• Domain authority scores - third-party estimates, not something Google uses

These look great in a deck, and they tell you almost nothing about whether the work is paying off.

Traffic is an indicator, not an outcome. At the end of the day, SEO is only working if it’s making you money.

Why “AI visibility tracking” isn’t the fix

The industry’s answer to all this has been a wave of tools claiming to track your AI visibility.

Be careful here.

AI answers don’t always trigger. Outputs vary by user and session. Models sample, compress and rewrite rather than cite consistently - a brand can be included in an AI answer one day, excluded the next, and re-included alongside different competitors the day after.

These tools don’t measure AI answers. They sample them.

And sampling is fine - genuinely useful, even - as long as we stop pretending it’s precision.

How to measure SEO performance now: eligibility, selection, outcome

Search no longer just ranks pages. It selects answers and recommendations - AI Overviews pick sources, ChatGPT picks brands to recommend, and Google’s interface picks which modules you appear inside.

So the framework I now use with clients models exactly that selection process, in three stages.

Stage 1: Eligibility

Before asking “do we show up?”, ask “how are we framed?”

Search your brand, ask the LLMs what your company is and who it’s for, and collect the descriptions that come back. Then trace where that framing comes from - your own site, reviews, forums, comparison posts.

Your entity is shaped more by what others say about you than what you say about yourself. That framing decides which answers you’re even eligible to appear in - and it’s why links and brand mentions still matter: they don’t just move rankings, they make you safer to recommend.

Stage 2: Selection (your Answer Share)

Take a cluster of the questions your buyers actually ask, run them repeatedly across Google’s AI results, ChatGPT and Perplexity, and score how often you’re included - and how: clearly recommended, listed among others, merely mentioned, or excluded.

That’s your Answer Share. It’s a sample rather than a precise number, but tracked over time it tells you whether you’re winning or losing the selection game in your category.

Stage 3: Outcome

AI impact often shows up before clicks do. So alongside the traditional KPIs, watch your branded search demand trend, ask “how did you hear about us?” on your forms (you’ll be surprised how often the answer is now ChatGPT or an AI answer), and connect rising Answer Share to enquiries and demos.

The only question that actually matters: did my brand show up when it mattered - and did business follow?

Reporting SEO to a board

One last thing, because it’s where most SEO reporting falls apart.

Boards don’t care about impressions, and they shouldn’t. When you report SEO, you’re not selling deliverables - you’re demonstrating value. Lead with the money: enquiries, pipeline, organic revenue, and what it would have cost to buy that traffic through ads.

If you want to put a number on it quickly, my free SEO ROI calculator does exactly that.

Final thoughts

SEO isn’t dying. It’s decoupling from the metrics we used to measure it.

If your dashboards don’t make sense anymore, it’s because they were built for a world that no longer exists - a world of stable SERPs, reliable volume data and rankings you could check like a league table.

Measure what’s close to money, treat visibility data as directional rather than precise, and start paying attention to how you’re framed and selected in AI search - because that’s increasingly where the game is decided.

And if you’d like help building a measurement setup your board actually understands, get in touch 🙂