The sentence that opens more and more of my calls: “We still rank first, and we're getting less traffic than before.” Sometimes it's traffic, sometimes it's leads, sometimes it's just a vague sense that the numbers on the SEO report and the numbers on the business dashboard have stopped agreeing with each other.
And the confusing part is that, on paper, nothing is wrong. Position one, right where it's supposed to be. The rank tracker is green. The problem is that the thing position was supposed to deliver — attention, visits, customers — is arriving in smaller and smaller amounts. This is, of course, largely a story about AI Search. But before I get there, I want to be honest about something: position was a leaky proxy long before anyone said “AI Overview.”
Position was always a proxy — we just stopped noticing
Nobody's actual goal was ever “rank first.” Ranking was a stand-in for what we really wanted: to be seen, to get the click, to win the customer. We optimized to position and reported on position because it was clean, countable, and easy to put in a slide. A single number that felt like the truth.
But it was always a proxy, and proxies drift from the thing they measure. For years the drift was small enough to ignore — top three got you most of the clicks, so position and outcome moved together. What's happened recently isn't that position stopped mattering. It's that the gap between the number and the outcome got too wide to keep pretending it isn't there.
This was never only an AI problem
Here's the part I want to make sure lands, because it gets lost in the AI hype: I was watching position lie about reality years before generative search existed. Two patterns came up again and again.
The first is CTR that collapses under a weak brand. You can hold a genuine number one and still lose most of the clicks to the brands ranking below you — because when people scan a results page, they don't read top to bottom, they look for a name they recognize and trust. I've seen a position-one result get a click-through rate that a strong brand would be embarrassed by at position four. The ranking was real. The authority to convert it into a click wasn't there. That's not an algorithm problem you can optimize your way out of — it's a brand problem wearing an SEO costume.
The second is the gap between what your rank tracker says and what's actually happening. Tools report a tidy “position 1,” but open Search Console and your real average position is lower — because the tracker measures a clean, personalization-free lab environment that fewer and fewer real users see. Actual results are shaped by location, history, and the brands a person is already connected to. Someone who keeps engaging with a bigger competitor gets that competitor pushed up and you pushed down, on the exact query where your tool swears you're first. The number in the report was never the number the user saw.
Before we blame AI for everything: position has been quietly overstating your visibility for a long time. AI just made the gap impossible to ignore.
Then AI removed the floor
What top-three used to guarantee was a floor — a minimum amount of attention that came with the rank. AI Overviews and AI Mode removed that floor. Now the answer often sits above your number-one result, complete enough that the user never scrolls to you. Your position didn't move. The attention attached to it did.
This is where the vanity-metric trap gets dangerous. Because position holds, the SEO report stays green, and everyone can keep pointing at a healthy-looking number while the business results quietly slide. You can even improve the metric — climb from three to one — and improve nothing real, because the click that used to come with position one is now being intercepted before it reaches anyone. Optimizing harder toward a broken proxy just gets you a more confident wrong answer.
What “visibility” actually means now
If position is no longer a trustworthy single number, the honest response isn't to find a new single number to replace it. It's to accept that visibility is now a basket:
- Presence in the answer — are you inside the AI Overview, the AI Mode response, the assistant's reply, not just under it?
- Share of AI citations — when the model answers questions in your category, how often is it you it names versus a competitor?
- Clicks and branded demand — real Search Console clicks, and whether people are increasingly searching for you by name.
- Assisted outcomes — leads and revenue that organic touched, even when it wasn't the last click.
And this is where brand quietly runs through everything. Your share of AI citations, your CTR at any given rank, your resistance to being personalized downward — all of it leans on whether the brand is recognized and trusted. That recognition is built outside the optimization itself, through the kind of brand work that isn't my job to do. What I can do is measure honestly where you stand and make sure a brand that has earned trust is structured to get credited for it. But no reporting dashboard can manufacture a recognition that was never built.
How I report it now
The practical change is in what I put in front of a client. Not a rankings table pretending to be a business result.
I segment by intent first — commercial queries and informational queries tell completely different stories now, and averaging them hides both. I lead with Search Console clicks and impressions, not average position, because clicks are closer to reality than a tracker's lab number. I add a view for AI-referral traffic — visits arriving from ChatGPT, Perplexity, and the like, tagged as their own segment, because that's a signal that didn't exist two years ago. And where it's relevant to the client's market, I track citation presence — how often the brand surfaces in AI answers versus competitors, using the LLM-visibility tools now entering the market.
The rankings table doesn't disappear. It just stops being the headline, because it stopped being the truth.
When position still matters
This isn't “ranking is dead.” A few places where position is still a fair proxy:
- Transactional and commercial queries. Where AI Overviews are thin and the click still follows the rank.
- Local and map-pack intent. Being in the pack still drives real calls and visits.
- Low-AIO niches. Industries the answer engines haven't reached yet. Watch them, but don't over-rebuild reporting that's still working.
- As a diagnostic, not a KPI. Position is still useful for spotting problems. It's just a bad thing to promise or celebrate on its own.
What to do Monday morning
You don't need to rebuild everything. You need to stop trusting one number.
- Audit your own reporting. How much of it is position, and how much is outcome? If a rankings table is the headline, you're measuring the proxy, not the result.
- Re-baseline on clicks and business metrics. Move average position down the page and put Search Console clicks, branded demand, and leads at the top.
- Add the AI view. Tag AI-referral traffic in analytics and check, by hand if you must, whether you're cited in the answers your category triggers.
- Segment by intent. Stop averaging commercial and informational queries into one meaningless line.
None of this makes ranking worthless. It just puts it back where it belongs — one diagnostic signal among several, instead of the number everyone agreed to trust long after it stopped telling the truth.
FAQ
Frequently asked questions
The most common cause today is that the results above your organic listing — an AI Overview or AI Mode answer — are satisfying the query before the user reaches you. Your position held; the attention that used to come with it didn't. Check whether your top queries now trigger an AI answer.
It's a useful diagnostic, but a poor headline KPI now. Position is a proxy for attention and revenue, and that proxy has drifted — you can hold a rank and lose the outcome. Report on clicks, branded demand, AI citations, and assisted conversions, and use position to diagnose rather than to celebrate.
Rank trackers measure a clean, un-personalized environment. Real users see results shaped by location, history, and their existing brand affinities. Someone who engages a lot with a competitor gets that competitor pushed up and you pushed down. Search Console reflects real, personalized positions, which is why it's the more honest number.
Yes — and it was common well before AI. The usual reason is weak brand recognition: on a results page, people gravitate to a name they trust, so a stronger brand ranking below you can take the click anyway. Ranking earns the placement; brand earns the click.
Not another single number — a basket. Presence in AI answers, share of AI citations versus competitors, Search Console clicks, branded search demand, and assisted leads or revenue. Segmented by intent, because commercial and informational queries now behave very differently.
No. It means the scoreboard changed. The work of earning visibility still works — but measuring it purely by ranking now hides more than it reveals. The teams that update how they measure will see clearly; the ones staring at a rankings table will keep being surprised by their own dashboards.
