"Traffic is down and it's probably the AI stuff" is a feeling, and your CFO runs on numbers. Closing that gap takes one afternoon of measurement. Industry-wide data already answers the big question: AI Overviews cut clicks by as much as 58% on the queries where they show up. They show up on roughly a quarter of queries, and overwhelmingly on informational ones.
None of those charts can tell you what's happening to your own SaaS, though. For that you need your own number, split by intent and priced in pipeline, with the other causes stripped out before it goes on a slide.
Step 1: Segment before you measure anything
Don't read a single trend until every query has a funnel label. Absorption hits some intents and spares others, and a blended chart averages that away. Export your Search Console queries and sort each one into a band:
| Band | What lands in it |
|---|---|
| TOFU-informational | definitions, how-to, glossary, what-is |
| MOFU-evaluative | comparisons, alternatives, vs-queries, integrations |
| BOFU-transactional | pricing, brand, trial, demo-adjacent |
Your labels don't need to be perfect, but they do need to be consistent. With a simple rulebook a junior can get through 500 queries in an hour. Don't skip it to eyeball the top twenty. Absorption hides in the queries you'd never think to check, and everything downstream stands on this step.
What you'll most likely see is the absorption signature, with informational clicks bleeding at stable positions while the evaluative and transactional bands hold. Your blended organic chart can't show that split, which is why teams argue about it for quarters. Once you can say which band is bleeding and at what position, the argument tends to be over by lunch.
Step 2: The CTR-delta math
This is the calculation that separates absorption from everything else. Pick two comparable windows and keep only the queries whose average position stayed stable across both, meaning within a position or so. Then work out how click-through rate changed in each intent band. If position held and CTR fell, that's absorption's fingerprint. A falling position is a different problem with a different checklist.
The number to write down is absorbed clicks per month = impressions × (old CTR − new CTR), summed over the affected queries. That's your loss, cleanly attributed. On a typical SaaS site it bunches up shockingly, and a dozen informational queries usually carry most of it. So instead of worrying about "our traffic," you're looking at "these twelve pages," and twelve pages fit on a to-do list. Put each query's absorbed volume beside it, since that ranking is also your fix order.
Built by the team that got cited in 48 hours.
Content generation, backlink building, AI visibility tracking, and Google rankings. One platform, zero guesswork.
Step 3: Price the absorption
The business runs on pipeline, so convert clicks into it. For each intent band, multiply the absorbed clicks by your historical value per organic session in that band. In every SaaS funnel we've seen, informational sessions convert at a fraction of the rate evaluative ones do.
Two things usually fall out of this. The informational losses, big as they look in clicks, turn out cheap in pipeline, because those sessions rarely converted directly. Their value was assist and awareness, and the citation layer keeps part of that. Evaluative losses are rarer, but they're real wherever Overviews now answer comparison-shaped queries, and they're expensive. That's where I'd put defensive effort first.
Add an asterisk before any of this goes on a slide. The absorbed informational clicks carried brand value your last-click model never credited, so what you've priced is a floor rather than the total. Say that out loud in the meeting and you'll be the most credible person in the room.
Step 4: Separate the confounders
Overview absorption shares a season with plenty of other things, so the number needs an alibi check before it ships. Lay Google's update calendar over your delta windows, including the August spam window and any core updates, and throw out any query whose position moved. Then open your own deploy log for the same period.
Seasonality is the last suspect. A year-over-year comparison on the bands Overviews didn't touch doubles as your control group. If BOFU held year over year while TOFU bled at a stable position, your attribution will stand up to the most skeptical VP you've got.
See your first AI citation report in under 5 minutes.
No setup calls. No onboarding meetings. Connect your domain and see where AI mentions your brand right now.
Step 5: Add the citation column
Clicks are half the measurement. The other half is what you got in exchange. For every affected query, note whether an Overview appears and whether you're cited inside it, because on absorbed queries being the source in the answer is the kind of visibility that's left.
Then cross the two ledgers. A query that was absorbed but cites you traded sessions for presence, and you can work with that. One that was absorbed without you lost both, and those are the real casualties. The citation-share trend on that second list becomes the KPI that stands in for the clicks you won't get back. Weekly per-engine tracking fills in this column for you, and it's the half of the instrument Search Console can't give you.
A worked model, with deliberately round numbers
This mid-market SaaS is hypothetical, with numbers picked for easy arithmetic rather than realism, so you can watch the model run end to end.
| Band | Impressions a month | CTR | Absorbed clicks a month | Value per session | Priced loss |
|---|---|---|---|---|---|
| Informational (80 queries) | 50,000 | Fell from 4.0% to 2.5%, position steady | 50,000 × 1.5% = 750 | $2 | $1,500 a month floor, roughly $18,000 annualized |
| Evaluative | 30,000 | Within noise, except two comparison queries that now show an Overview | 400 combined, from those two | $9 | $3,600 monthly exposure |
Most of the headline loss sits in a small corner of the informational row, where 11 of the 80 queries carry over two thirds of the absorbed volume. That makes your working list 11 pages. The evaluative row looks calm by comparison, yet those two comparison queries alone outprice the entire informational bleed.
Read the two rows together and the plan changes. The loss everyone notices is informational and cheap, while the quiet evaluative one is small and expensive. So the defensive budget goes to two comparison pages, plus citation share on the eleven, and that looks nothing like "traffic is down, write more content."
Put your own numbers through the same frame and expect the same shape: big, cheap losses at the top of the funnel, rare and costly ones in the middle, and a to-do list much shorter than the panic suggested.
The instrumentation, all of it cheap
You don't need enterprise tooling for any of this. Search Console exports and a spreadsheet cover the segmentation and the CTR-delta math, and the formulas take twenty minutes the first time before you reuse them forever. A rank tracker with SERP-feature detection handles Overview presence and the cited-or-absent flag for each query.
Search Console's own generative-AI view adds directional impressions, though it shows no clicks and no query split. The citation column runs on a weekly fixed panel per engine, which you can keep by hand in a spreadsheet or automate in tracking. The whole stack costs less than one month of the confusion it clears up. What really cost teams was that nobody owned the afternoon.
Five mistakes that invalidate the number
Each of these quietly breaks the attribution. Two are about timing, so keep the update calendar open while you set your windows.
| Mistake | Why it breaks the number | Fix |
|---|---|---|
| Keeping unstable-position queries | The CTR delta measures the ranking change, and the attribution collapses | Filter them out |
| Windows straddling an update | A core or spam update inside the window imports its noise | Cut windows at the calendar's boundaries |
| Drifting intent labels | Relabeling between runs makes trends unreadable | Freeze labels, log changes, re-baseline if you must recategorize |
| Single-week windows | Overview coverage flickers while Google tests, so you'd measure the flicker | Four-week windows minimum |
| Presence without position | Source nine in an Overview is presence in name only, and position multiplies value several-fold | Track where you sit, never just whether you appear |
The slide that ends the argument
All five steps collapse into one slide, which turns a quarterly argument into a decision. Open with the intent-segmented click trend so the room can see where the bleed lives, and put the absorbed-click volume and its priced floor beside it. A one-line confounder note names your control band. The citation cross comes next, cited versus absent on the affected queries.
The last line is the reallocation, and it's the reason you measured in the first place. Informational pages get regraded on presence and any evaluative losses get defensive depth, while the content capacity you free up goes to decision-stage and citation-earning work. The impact number exists to justify moving effort to where clicks still live and citations get decided.
Re-run the full model every quarter, and again after anything that reshapes the results page, whether that's an update, a visible expansion in Overview coverage or a model release that moves your citation columns. In between, weekly tracking keeps watch, and each quarterly run recalibrates the prices and the labels. Impact models go stale as fast as the surface they measure, and right now no surface in marketing moves faster than this one.
Your first run takes an afternoon, and once the tracking runs, the model maintains itself weekly. The other option is another quarter of "it's probably the AI stuff." That costs more than every tool in this workflow put together, and you pay it in decisions nobody made and budgets defended with vibes.
Your competitors are building backlinks while you read this.
Organic outreach, social mentions, and citations from Editorial Circle member sites, with managed backlinks available as an add-on. Grow your domain authority without running the campaign yourself.
