The short answerAn SEO forecast estimates the organic clicks and revenue your pages could add over a period. For B2B SaaS, multiply monthly searches by click-through rate at your target position, discount searches showing an AI Overview, then carry clicks through your own lead, SQL and contract-value rates to pipeline. It is a range with written inputs, not a promise.
Most forecasts in an agency proposal are one line of arithmetic dressed up as a chart. The arithmetic is fine. The inputs are where forecasts go wrong, and in 2026 the click-through curve and the Search Console data behind it are both worse than most people assume.
What does an SEO forecast look like on one page?
Two lines carry the whole method:
Added clicks = (searches without an AI Overview × CTR at the target position) + (searches with one × CTR at the target position × (1 − AI Overview cut)) − clicks you get today
Added pipeline = added clicks × visit-to-lead rate × lead-to-SQL rate × average contract value
The table runs that formula on one worked example. Every input below is an illustrative assumption you replace with your own numbers: a cluster of 20 non-brand keywords with 10,000 combined US searches a month, half of those searches on results pages with an AI Overview, 10 clicks a month today from page two, a 2% visit-to-lead rate, 25% of leads reaching SQL stage and a $20,000 average contract value. The click-through rates are real: Backlinko's position curve for results without an AI Overview, cut by the per-position drop Ahrefs measured in December 2025 for results with one. We use Backlinko's curve because it's a curve many forecasting templates still use, so the table shows what a proposal built on it would claim. Its data predates AI Overviews and sits far above Ahrefs' 2025 position-one reading, so treat every column as an upper bound. The next section shows how to swap in a curve of your own.
| Illustrative example (assumed inputs) | Baseline (do nothing) | Conservative | Expected | Aggressive |
|---|---|---|---|---|
| Average position reached | Page two | 8 | 4 | 2 |
| CTR with no AI Overview (per Backlinko) | n/a | 3.3% | 8.4% | 15.8% |
| AI Overview cut at that position (per Ahrefs) | n/a | -28.8% | -38.8% | -50.8% |
| CTR with an AI Overview | n/a | 2.35% | 5.14% | 7.77% |
| Monthly clicks (5,000 searches at each of the two CTRs above) | 10 | 282 | 677 | 1,179 |
| Clicks gained a month | 0 | 272 | 667 | 1,169 |
| Leads gained a month (2%) | 0 | 5.4 | 13.3 | 23.4 |
| Leads at SQL stage, gained a month (25%) | 0 | 1.4 | 3.3 | 5.8 |
| Pipeline gained a month ($20,000 contract value) | $0 | $27,200 | $66,700 | $116,900 |
The expected column, worked through: 5,000 searches × 8.4% = 420 clicks. The other 5,000 searches see an AI Overview, so 8.4% × (1 − 38.8%) = 5.14%, and 5,000 × 5.14% = 257 clicks. Then 420 + 257 = 677, minus the 10 you already get, leaves 667 added clicks. 667 × 2% = 13.34 leads, × 25% = 3.335 at SQL stage, × $20,000 = $66,700 of pipeline a month once the rankings hold. Each pipeline figure is worked from the rounded click count, and the lead rows are rounded to one decimal.
On the three page-one queries we could test against our own Search Console data, the Backlinko curve predicted about 25 clicks for the quarter and we got none. Three queries can't give you an average miss, but they show how far the curve can miss on a page with an AI Overview.
There are two other methods. A top-down forecast fits a time-series model to your traffic history and projects it forward. A baseline (tools call it inertial traffic) shows what happens if you do nothing, and the bottom-up uplift sits on top of it.
Which click-through curve should you plug in?
The curve is the biggest lever in the forecast, and public curves disagree by a factor of several.
Start with Backlinko's study. Its title says it analysed 4 million Google results, built on CTR data across 1,312,881 pages and 12,166,560 search queries, and its headline finding is that the number one organic result has an average CTR of 27.6%. The per-position values only appear in a chart image: 27.6, 15.8, 11.0, 8.4, 6.3, 4.9, 3.9, 3.3, 2.7 and 2.4 per cent for positions one to ten. The study was first published in 2019 and the chart was uploaded in 2022. AI Overviews weren't in the data.
Ahrefs' own Search Console sample tells a very different story for informational keywords on desktop. Position-one CTR was 7.6% in December 2023 and 3.9% in December 2025, and that's for keywords without an AI Overview. Same position, a fraction of the clicks.
Both curves are honest about their samples. A 2019 all-query curve tells you what a page could earn before AI Overviews, so use it as the ceiling it is. Forecasting 2026 informational content on it alone, which is what a template copied from an old guide does, overstates the clicks.
Three rules hold up:
- Build a custom curve from your own Search Console clicks and positions if you have enough data. Ahrefs recommends this because generic curves "may not be representative of your site."
- If you borrow a curve, borrow one that splits results with and without an AI Overview. Advanced Web Ranking publishes one from Search Console data, refreshed every month since 2015, and advises re-forecasting with the AI Overview curve instead of the blended average.
- Write the curve's name and date in the forecast. If an agency can't tell you which curve it used, the forecast has no foundation.
How much should you discount for AI Overviews?
Start from the per-position cut Ahrefs measured, which reaches 58% at position one, and recheck it every month. Ahrefs and Seer point in different directions, and that disagreement is the useful part.
Ahrefs compared the desktop click-through rates of 300,000 keywords in December 2023 and December 2025. Position-one CTR on keywords that gained an AI Overview fell from 7.3% to 1.6%. Informational keywords without one fell from 7.6% to 3.9% over the same period. Net of that general decline, Ahrefs puts the AI Overview's effect at a 58% lower CTR for the top-ranking page. The cut shrinks down the page: -50.8% at position two, -46.4% at three, -38.8% at four, down to -19.4% at ten. Its earlier run, March 2024 against March 2025, measured a 34.5% cut. So the gap widened between the two runs.
Seer Interactive's newer study points the other way. Across 53 brands, 5.47M queries and 2.43B organic impressions from January 2025 to February 2026, organic CTR on queries with an AI Overview climbed from a floor of 1.3% in December 2025 to 2.4% in February 2026. Being cited inside the AI Overview delivered 120% more organic clicks per impression than not being cited, yet on informational queries a cited page still underperformed queries with no AI Overview by 38%. Seer's figures are query-level CTRs across all positions, so they can't stand in for a position-one rate. Seer also reports that its own regression model "missed organic search CTR by 0.6 to 2.6 percentage points across every segment," and advises wider error bars and more frequent check-ins for organic.
Pew Research Center measured the same effect from the user's side. In March 2025 browsing data from 900 U.S. adults, people who saw an AI summary clicked a traditional result in 8% of visits, against 15% when there was no summary. They clicked a link inside the summary in just 1% of visits.
Our position: default to the December 2025 cut from Ahrefs, the harsher reading (the table above does), show the same forecast with a cut nearer the earlier 34.5%, and re-run it monthly against what Search Console records. A single multiplier applied for twelve months is a guess with a decimal point.
How often do AI Overviews appear, and why do the numbers disagree?
You only discount the keywords that trigger an AI Overview, so prevalence matters. The published figures look contradictory until you check the unit each one counts.
| Source | What it counts | Figure |
|---|---|---|
| Semrush, 10M+ keywords, 2025 | Share of tracked keywords | 6.49% in January, 24.61% in July, 15.69% in November |
| Pew Research Center, March 2025 | Share of real U.S. searches | 18% |
| Seer Interactive, 2025 to 2026 | Share of queries by intent | 36% informational, 8% commercial, 5% transactional |
| Seer Interactive, 2025 to 2026 | Share of informational queries, by shape | 95.4% of comparison queries, 85.9% of question queries |
Keywords, searches and intent segments are different denominators, so never average them. Check the results page for each target keyword, or pull SERP features from your rank tracker, and tag each keyword yes or no. If your list is mostly informational comparison and question keywords, expect most of it to carry an AI Overview.
What did our own Search Console data say about forecast inputs?
In a sample of 56 of our highest-impression non-brand queries, 32 logged more Search Console impressions in the US in Q3 2026 than the estimated US searches for the whole quarter (DataForSEO Labs), and those 32 carried 81.3% of the sample's US impressions. A real searcher can't produce more impressions than there were searches. Most of those impressions weren't people.
That matters for forecasting because impressions and average position are the inputs most forecasts are calibrated on. Thousands of impressions at position 15 look like demand waiting to be captured. Some of it isn't demand.
The detail, from loudface.co's own data:
- Where impressions exceeded searches, the median ratio was about 4x, and 21 queries logged 3x or more.
- The most extreme was "query fan out chatgpt": 965 US impressions against 40 estimated US searches, a ratio of 24.1x, at an average position of 18.1.
- The 25 queries averaging positions 11 to 20 logged 9,543 US impressions against 5,540 estimated US searches (1.7x), with 1 click between them.
- The 28 queries at position 21 or lower logged 5,962 US impressions against 24,100 estimated searches, fewer impressions than searches, as you'd expect that deep. Band by band, most of the excess sits on positions 11 to 20, though some individual queries below position 20 still ran over.
Google defines an impression as a link that appeared on the current results page "whether or not the item is scrolled into view." A rank tracker that loads deep results pages therefore creates impressions. When Google stopped honouring the results-per-page parameter (num=100), discovered by the SEO community on 12 September 2025, an analysis of about 320 Search Console properties reported by Search Engine Roundtable found 87.7% of sites saw impressions drop and 77.6% lost unique ranking keywords. Google said only that the parameter "is not something that we formally support." It never confirmed scrapers caused the impressions. Our Q3 2026 sample sits after that change, so the change didn't remove all of the non-human reads.
Only 3 of our 56 queries averaged a page-one position, between 4.6 and 6.3. Backlinko's curve applied to their estimated searches predicted 25.2 clicks for the quarter. Search Console recorded 0 US clicks on them. All three results pages showed an AI Overview in the latest snapshot DataForSEO held when we read it on 1 November 2026, not in a check we made during the quarter. Three queries is far too small a sample to estimate a CTR, so don't read it as an average miss. It shows how far a pre-AI-Overview curve can miss on a page that has one. Across the whole sample, 29 of the 56 queries (52%) showed an AI Overview in DataForSEO's latest snapshots and 13 had no usable snapshot, so at least half of the queries in our own forecast would take the discount.
Method. We started from the 400 non-brand queries with the most worldwide Google Search Console impressions between 1 July and 30 September 2026, dropped the 69 longer than 10 words, and looked up the remaining 331 in DataForSEO Labs for the US. DataForSEO returned data for 70 of them, and we kept the 56 that had a full US search volume for July to September and weren't a client's brand name. Picking the highest-impression queries tilts the sample toward queries whose impressions run above their searches, so read the 32 of 56 as a finding about this sample, not about every query we appear for. We compared quarterly US impressions with quarterly estimated US searches, then applied Backlinko's published curve to the page-one queries. DataForSEO volumes are modelled estimates and coarse at low volume (most of the high-ratio queries sit at 10 to 40 searches a quarter), so read the direction as the finding and the exact multiple as rough.
What to do with it: before you calibrate on impressions, compare each query's impressions with its search volume, and where impressions run higher, use searches. Treat average position on low-volume queries as noise too, because a link must get an impression for its position to be recorded.
How far can you trust search volume, ranking time and seasonality?
Search volume. Ahrefs compared Google Keyword Planner volumes with Search Console impressions in 2021, for 72,635 keywords in the 1K to 10K range, and found 54.28% of Keyword Planner volumes were overestimates, while 45.22% were roughly accurate (within 50% of impressions). Keyword Planner groups close variants and rounds volumes into buckets, and Google says its default figure is averaged over 12 months. Ahrefs rated its own volumes roughly accurate in 60% of cases, which is a vendor grading itself. Use volume for ranking keywords against each other, and be wary of multiplying it by anything precise.
Time to rank. Ahrefs tracked new pages for a year: only 1.74% of newly published pages reached the top 10 within a year, down from 5.7% in 2017. A cleaner cut of non-empty English pages put it at 6.11%. Of the pages that did make the top 10, 40.82% got there within a month, and 72.9% of pages in Google's top 10 are more than three years old. Ahrefs suggests revisiting content that still isn't in the top 10 after about six months. A forecast showing full traffic in month two on a young domain ignores all of this. Our guide on how long SEO takes to work covers the timing in more depth.
Seasonality. Search Console keeps 16 months of data, and Prophet, the open-source model behind many top-down forecasts, only fits yearly seasonality by default when the series is more than two cycles long. Sixteen months is less than two years, so a default Prophet run on Search Console history alone won't model your seasonal dip. Use a longer series from Google Analytics or a bulk export. Google Trends won't fill the gap with volumes either: it scales relative interest from 0 to 100.
How do you turn forecast clicks into pipeline?
With your own funnel rates. There's no primary, non-agency benchmark for organic B2B SaaS visitor-to-lead or lead-to-SQL conversion. The most-quoted SaaS figure, Unbounce's 3.8% median, is for landing pages across channels, not an organic blog visit.
So pull three rates from your own GA4 and CRM, for organic traffic only:
- Visit-to-lead (form fill, demo request or signup, whichever your sales team counts).
- Lead-to-SQL or lead-to-opportunity.
- Average first-year contract value, and if you want revenue instead of pipeline, your win rate.
Then add the sales-cycle lag: if your cycle runs a quarter, pipeline from a March ranking closes in the summer.
The forecast ends at pipeline. A forecast that stops at clicks or impressions lets everyone celebrate a chart while revenue stays flat, and our back-test shows impressions alone can mislead. If an agency's forecast has no pipeline line, ask it to add one, and to say which of your rates it used. Our guide to measuring AEO agency ROI covers the reporting side, and the ROI math on SEO, AEO and CRO weighs which channel should get your next marketing dollar.
Forecast branded and non-branded traffic separately. Branded searches come from people who already know you, and they move with sales, PR and paid spend. Within branded, only new queries count as SEO lift: people searching your login or account pages are customers, and that traffic is noise.
How do you check an agency's SEO forecast before you sign?
Ask these nine questions. A strong agency answers each one in a sentence and shows you the spreadsheet.
| # | Question to ask | Strong answer | Red flag |
|---|---|---|---|
| 1 | Which CTR curve did you use, and how did you discount AI Overviews? | Names the curve and its date, splits keywords with and without an AI Overview, and applies a stated cut | "Industry average CTR" with no source, or a 2019 curve applied to every keyword |
| 2 | What happens to our traffic if we do nothing? | A baseline line under the forecast, so the uplift is shown on its own | Only one line, starting from zero |
| 3 | Did you separate branded from non-branded traffic? | Non-brand forecast on its own; branded lift counted only for new branded queries | Branded traffic counted as SEO growth |
| 4 | How did you clean the Search Console data? | Compared impressions with search volume, used searches where impressions run higher, and flagged low-volume positions | Uses raw impressions and average position as they are |
| 5 | What does the ramp look like by month? | A month-by-month curve with ranking time stated per keyword group | Full traffic from month two |
| 6 | How do clicks become pipeline? | Uses your own visit-to-lead, lead-to-SQL and contract-value rates, with the sales-cycle lag | A borrowed conversion rate, or no pipeline line at all |
| 7 | What's the range? | Three scenarios (conservative, expected, aggressive) with the inputs behind each | A single number |
| 8 | How often will you report forecast against actual? | Monthly, on the same sheet, with the curve and discount updated as data arrives | Reports that switch metrics when the forecast misses |
| 9 | What happens if we miss the forecast by a wide margin? | Names the review trigger, how the plan changes, and what it learns from the miss | No answer, or a promise that it won't happen |
One answer ends the conversation on its own: a guarantee. "No one can guarantee a #1 ranking on Google," says Google's own guidance, which adds that an SEO audit should give "realistic estimates of improvement." A forecast is an estimate. An agency that guarantees a traffic number has either padded the inputs or plans to hit the number with keywords that will never produce a lead.
For the wider version of this vetting, our audit of how AI search agencies report pipeline checks which of them can name a client whose leads came from AI search.
Should you forecast AI search visibility too?
Yes, but as a range per engine, never as a click count or a citation count. AI answers change from run to run, so a single reading tells you very little.
SparkToro had 600 volunteers run 12 prompts through ChatGPT, Claude and Google's AI a combined 2,961 times. For ChatGPT or Google's AI, there was less than a 1 in 100 chance of getting the same list of brands in any two responses, Claude was only slightly more likely to repeat a list, and it took about 1 in 1,000 runs to see two lists in the same order. Its conclusion still supports measurement: the percentage of answers that mention you is a reasonable way to read how prominent you are.
Our own study of 1,636 AI answers, collected from 5 to 18 October 2026 across 39 B2B agency questions, found ChatGPT gave an identical list on consecutive days just 2.9% of the time (14 of 488 day pairs), Perplexity 20.3% and Google AI Overviews 5.8%. A single check missed an agency named on at least half of a question's days 26.8% of the time on ChatGPT.
So the AI line in a forecast should look like this:
- Metric: visibility, the share of AI responses that mention your brand, read per engine (ChatGPT, Perplexity, Google AI Overviews) over a week or a month. Our explainer on what AI visibility is covers how each engine is measured.
- Shape: a starting reading and a range for each engine.
- Timing: AI citations move at three speeds. First pickup can happen within hours to a day, a consistent slot in the answers takes weeks, and dominant share, with branded lift and pipeline, takes months. Google AI Overviews updates fastest, Perplexity moves daily to weekly, and ChatGPT is slower to pick up new pages.
That makes the AI Overview citation the earliest signal in the whole forecast. It also feeds back into the click line, since Seer measured a cited brand earning 120% more organic clicks per impression than an uncited one.
Which tools build SEO forecasts?
You don't need a paid tool. A spreadsheet with your Search Console export, a dated curve and your CRM rates beats a polished chart built on defaults. The tools save time:
| Tool | What it does for forecasting |
|---|---|
| Ahrefs free Colab notebooks | Prophet time-series forecasts of traffic, traffic value, competitors and pages, 365 days ahead by default, with an 80% uncertainty band. Ahrefs also offers a Traffic Potential metric |
| Advanced Web Ranking | Linear or exponential projections for up to 24 months, with monthly updates comparing actual traffic to predicted and inertial traffic |
| SEOmonitor | Baseline (inertial) traffic and SERP-adjusted CTR curves that account for features such as AI Overviews |
| Semrush Enterprise | Scenario modelling and revenue prediction; its page gives no method detail or price |
Prophet is now in maintenance mode, with no new features planned. And Advanced Web Ranking says it plainly on its own product page: "SEO forecasts can never be truly precise." Patrick Stox at Ahrefs adds that more complex forecasts "aren't necessarily better."
What a forecast is for
A forecast earns its place in a budget meeting when every assumption is visible and it gets checked against reality monthly. The number on the slide matters less than knowing why it moved.
If you're comparing agency proposals, run the nine questions above against each forecast. The one that shows its inputs and ends at pipeline is the one worth a second meeting. Our own engagements start from $5k a month, and we report AI visibility per engine against signups, booked demos and captured leads from the first month, which is also our busiest one. See our pricing for the tiers, or our page on SEO for SaaS companies.
