TL;DR
- OrganikPI measured cost per click and search result structure for 160 commercial queries across eight verticals, capturing the identical panel twice in 2026.
- Median cost per click ranged from $1.49 in retail to $53.95 in B2B SaaS, a 36x spread.
- AI Overview coverage rose from 72.5% in July to 80.6% in August.
- Total cited-source slots fell 12.5% even as AI Overviews increased, and the median answer kept only 40% of the sources it cited six weeks earlier.
There is no single answer to SEO versus Google Ads, and our data shows why. We measured the cost of a click and the shape of the search result for 160 commercial queries across eight industries. A click costs $1.49 in retail and $53.95 in B2B SaaS, a gap of 36 times, so the same budget buys completely different outcomes depending on what you sell.
We also ran the identical 160 queries twice, six weeks apart, and watched what Google did to them. AI Overview coverage on those commercial searches rose from 72.5% to 80.6%. The median answer now cites 7 sources where it cited 9 in July. Of the 110 queries that kept an AI Overview across both captures, not one kept the same set of sources.
What we measured, and how
We started from the query panel built for our earlier study on ads inside Google AI Mode: 160 commercial and transactional searches, 20 in each of eight verticals, captured on Google United States desktop through the DataForSEO SERP API with asynchronous AI Overview content enabled.
For this study we added two things. First, the money: cost per click, search volume, and top-of-page bid ranges from Google Ads, plus keyword difficulty from Semrush. Second, time: we re-ran every one of the 160 searches on 20 August 2026, six weeks after the first capture, so any change is a change in Google’s behaviour rather than a different sample.
One gap is worth naming up front. Google Ads reported no cost data for 11 of the 160 keywords, concentrated in healthcare and legal, so cost-per-click medians are taken over the keywords that returned a price: 15 of 20 in healthcare, 17 in legal, 18 in education, 19 in B2B SaaS, and all 20 in the other four verticals. Everything else uses the full 160.
Every keyword-level number is published. The full dataset is linked at the end of this article, and the calculator below runs on the same figures.
The 36x spread is the real finding
Most SEO versus PPC advice is written as though every business faces the same trade. Our numbers say otherwise. Median cost per click ran from $1.49 in retail and ecommerce to $53.95 in B2B SaaS, with legal at $33.36 and insurance and finance at $26.45.
Individual keywords go further. “Best applicant tracking system” carries a $205.06 cost per click on 590 searches a month. “Best hr software” runs $150.48. At those prices a modest paid programme costs more in a quarter than a year of content work, which is the arithmetic that makes organic investment in B2B SaaS worth doing even when rankings are slow.
The reverse holds in retail. A $1.49 click is cheaper than almost any content programme can deliver once you count upkeep, which is why ecommerce search strategy tends to reward paid coverage and organic depth on a narrower set of pages.
Google added AI Overviews to 13 more of our queries in six weeks
On the same 160 searches, AI Overview coverage moved from 72.5% in July to 80.6% in August. Insurance and finance moved the most, from 70% to 95%. B2B SaaS reached 100%. Legal did not move at all, holding at 45%, which matches the pattern we described in our guide to GEO in regulated industries.
Six queries lost an AI Overview and 19 gained one, so churn touched 25 of the 160. The direction is clear enough on a single interval, but one six-week gap is one observation. Treat the slope as directional and the endpoints as measured.
The average first organic result sat at position 2.05 in B2B SaaS and 4.10 in legal. On most of these searches a person meets an AI answer before they meet an independent link, which is the mechanic behind zero-click search and the reason AI Overviews change click behaviour.
More AI answers, fewer places to be cited
This is the part that surprised us. August produced 13 more AI Overviews than July across the panel, but 133 fewer cited-source slots. Total slots fell from 1,061 to 928, a drop of 12.5%, because the median answer now cites 7 domains where it cited 9.
We checked that this is not an artefact of how we fetched the answers. Google returns some AI Overviews already rendered and loads others asynchronously, and the mix shifted between our captures. Holding that constant, on the 73 queries that came back rendered in both captures, the median still fell from 9 sources to 7. The answer text itself also got shorter, from a median of 1,482 characters to 1,279, and text is parsed separately from the citations.
Coverage is expanding and the guest list is getting shorter at the same time. If you are competing for AI citations, the number of available seats fell while the number of tables grew. That reframes what an AI Overview is worth compared with a featured snippet.
Citations do not stay won
We compared the cited domains for the 110 queries that held an AI Overview in both captures. The median answer kept 40% of the sources it cited six weeks earlier. Half the sources in a typical August answer were new. Three queries shared no sources at all with their July version, and none of the 110 kept an identical set.
Retention was consistent across all eight verticals, from 25.0% in legal to 43.7% in healthcare, so this is not one category behaving oddly. A citation is closer to a rented position than an earned one, which makes tracking citation velocity more useful than checking a ranking once a quarter, and makes a sudden citation drop a normal event rather than a crisis.
The domains on the receiving end were familiar. YouTube appeared in 67.4% of August’s AI Overviews, up from 58.6% in July, and Reddit in 37.2%. That is the same concentration we covered in why YouTube and Reddit dominate AI citations.
Work out your own number
The calculator below is seeded with the measured figures for each vertical: median cost per click, keyword difficulty, AI Overview rate, and how far down the page the first organic result sits. You supply the parts only you know, which are how many clicks you want, what the content route would cost to build, what it costs to hold, and how long you think it takes to rank.
It applies one adjustment worth explaining. Pew Research found people clicked a search result on 8% of visits when an AI summary was present and 15% when it was not. Combined with the AI Overview rate we measured in each vertical, that gives the share of its pre-AI clicks a ranking still earns. In B2B SaaS, where every query in our August capture returned an AI Overview, that works out at 53%.
Open the Search Budget Calculator and change the inputs. It is free, needs no signup, and nothing you type leaves your browser.
Two results are worth trying. In retail, at $1.49 a click and a realistic upkeep budget, organic never overtakes paid, because holding the rankings costs more per click than buying them. In B2B SaaS, at $53.95 a click, it overtakes inside about a year and a half even after a twelve month wait to rank. Same calculator, same assumptions, opposite answers to the question of whether SEO is worth funding. The vertical decides that, not the strategy.
Where the standard numbers get repeated without their footnotes
Two things came up while we were sourcing this piece that are worth passing on.
The widely quoted figure that only 1.74% of new pages reach the top ten within a year comes from Ahrefs, published 15 May 2025. The same study says that when you filter to English pages with actual content, 6.11% get there, and adds that this “might be a more fair number to use.” The cautious version travels; the correction usually does not.
Keyword difficulty also depends on who you ask. Across the 157 keywords where both tools returned a score, DataForSEO and Semrush disagreed by a median of 24 points, with Semrush higher on 139 of them. Neither is wrong, because they model different things, but a build-versus-buy call made on a difficulty score of 25 reads very differently at 49. We used Semrush difficulty throughout this study and say so wherever it appears.
How to read this for your own budget
- Check your click price before your channel philosophy. A $1.49 click and a $53.95 click are different businesses. Everything else follows from that.
- Discount organic forecasts for the AI answer. On commercial searches in our panel, 80.6% now return an AI Overview, and the first organic link is rarely first on the page.
- Budget for holding a citation, not winning one. Median retention across six weeks was 40%.
- Expect nothing for the first year. Ahrefs puts the average page ranking first at five years old, with 72.9% of top ten pages older than three years.
- Do not fund a paid programme you cannot exit. Paid stops the day you stop paying, which is the trade our GEO and SEO budget split looks at from the other direction.
If you are putting a number to any of this internally, our guides to measuring revenue from AI search traffic and what GEO actually costs cover the parts this calculator deliberately leaves to you.
Limits of this study
- Commercial queries by design. The 80.6% AI Overview rate is high because we sampled high-intent commercial searches. It is not an all-query rate. Pew found 18% across a general sample of 68,879 searches.
- Two captures, not a time series. Six weeks is one interval. The churn figures describe the distance between two points, not a rate you can annualise.
- Twenty keywords per vertical. Vertical splits are directional. The 160-keyword totals are the firmer numbers.
- Automated, logged out, United States desktop. Ads and personalisation differ for a signed-in person, and we captured no paid results at all through this method.
- No conversion rates. We found no study that benchmarks organic against paid conversion on a shared methodology, so the calculator compares the cost of a click and stops there.
Get the data
The full keyword-level dataset is published as a CSV: 160 rows covering cost per click, search volume, top-of-page bid range, Semrush difficulty, AI Overview presence in both captures, cited-domain counts, source retention, and first organic position. Use it, check it, argue with it.
Download the dataset (CSV, 160 rows)
If you want the method behind the SERP capture rather than the results, it is documented in the original citation squeeze study. For the content side of getting cited at all, start with why calculator and tool pages earn AI citations.