How to Do SEO Forecasting in the Age of AI Overviews

SEO forecasting is the practice of using historical performance, keyword data, and expected click-through rates to predict a site’s future organic traffic, leads, and revenue. The method still works. But the click-through rates it leans on have dropped hard on any search where an AI Overview shows, so a forecast built on the old numbers will overpromise, and overpromising is the fastest way to lose a client you just won.

You’ve probably felt this already.

Rankings hold steady, or even improve, and the traffic doesn’t follow. That gap is the whole reason forecasting is harder now than it used to be, and it’s also why a forecast that accounts for it is worth more to a client than one that doesn’t. This guide is the hands-on version: the exact tools, steps, and spreadsheet formulas to build a forecast you can defend, with a CTR model that reflects what actually happens on the page today.

What Is SEO Forecasting?

SEO forecasting is the process of estimating a site’s future organic performance from data you already have. You take search volume, current rankings, click-through rates, and past traffic, then project forward to a number you can put in front of a client or a finance team.

The naive version looks like this. Grab the search volume for a keyword, assume you’ll rank number one, multiply by a fixed click-through rate, and call the result “traffic.” That math is where most bad forecasts come from. It assumes every number-one ranking earns the same share of clicks, and that’s no longer true, because a growing share of searches now resolve on the results page itself.

A useful forecast does the opposite. It treats the click-through rate as the variable that moves, segments keywords by what the results page actually looks like, and hands the client a range instead of a single confident number.

Why SEO Forecasting Still Matters, and What AI Overviews Changed

Forecasting earns its keep in three places: winning the pitch, setting expectations you can hit, and justifying the budget. A forecast turns “trust us, SEO works” into a roadmap with numbers attached, which is a much easier thing for a skeptical prospect to say yes to. It also protects you later, because a client who agreed to a realistic projection won’t be surprised in month four.

Here’s what changed.

On searches where Google shows an AI Overview, the answer often sits at the top of the page and the user never clicks through. The scale of that isn’t small. Seer Interactive analyzed 3,119 informational queries and found organic click-through rate on AI Overview searches fell 61%, from 1.76% to 0.61%, and you can read the full study here. Study after study has pointed the same direction. Rankings became a weaker predictor of clicks, and clicks stayed the thing that drives leads.

So the old forecasting model didn’t die. It just picked up a discount that most forecasts ignore.

Two practical consequences for your work:

• A number-one ranking on an AI Overview search is worth a fraction of what it used to be. Forecast it as if it were mid-page, not top-of-page, unless your brand is the one being cited in the Overview.
• Being cited inside the AI Overview matters more than the blue-link position. Brands cited in AI Overviews earned about 35% more organic clicks than those that weren’t, which means citation presence belongs in your forecast as its own input.

What You’ll Need

You can build a solid forecast with five inputs, and most of them are free. Start with the data, then add the layer that ties it into a client report.

• Google Analytics 4 and Google Search Console. Your first-party record of what already happens: organic sessions, conversions, impressions, clicks, and average position. Audit the setup before you trust it. Broken conversion tracking is the most common reason a forecast is wrong from the first cell.
• At least twelve months of historical traffic. Twelve months lets you see seasonality instead of mistaking a summer dip for a decline. Less than that and you’re guessing at the shape of the year.
• A keyword tool with click and volume data. Semrush or SE Ranking both do this, and both now flag AI Overview presence, which you’ll need for the discount step.
• A spreadsheet. Google Sheets is where you’ll model scenarios and show your working. Nothing fancy required.
• A reporting layer to put it in front of the client. This is where Swydo’s SEO reporting tool fits. It pulls your forecast and your live SEO metrics into one client-facing report, so the projection and the actual results sit side by side instead of in two files nobody opens together. The 14-day trial covers 10 data sources, which is enough to wire up a full client before you decide.

How to Forecast SEO Traffic, Step by Step

The workflow is the same whether you’re forecasting one page or a whole account: set the goal, pull a clean baseline, flag the keywords AI Overviews touch, then model the traffic and wrap it in a range. Steps one and two are prep. Steps three through five are three ways to do the actual modeling, and you can use one or stack them.

1. Set the Goal and Pull Your Baseline

Start with the metric the client actually cares about, not the one that’s easiest to project. Some want leads, some want branded demand, some just need a number to defend their own budget upward. Pin down which conversions count and what “good” looks like at three, six, and twelve months, and decide which SEO metrics you’ll report on, because a forecast is only useful if you track the same thing you predicted.

Then build the baseline:

1. Pull twelve months of organic sessions and conversions from GA4, and impressions, clicks, and average position from Search Console. Line them up by month.
2. Audit the tracking. Check conversion events and goal setups so you’re not projecting off broken data.
3. Note the seasonal peaks and any past algorithm-update dips, so you can tell a normal dip from a trend.
4. Pull the high-impression, low-click keywords from Search Console. You already rank for these, so they’re the fastest wins to fold into a forecast.

2. Flag the Keywords That Trigger AI Overviews

Split your keyword set into two lists before you model anything: searches where an AI Overview appears, and searches where it doesn’t. The two lists get different click-through rates, and lumping them together is what produces the confident, wrong number.

Both major keyword tools surface this. In Swydo, the Semrush integration pulls two AI Overview fields into a report: the count of keywords triggering an AI Overview, and the exact keywords triggering placement. That shows you, per client, how much of their target set is exposed to click compression, and whether the brand is being cited in those Overviews over time. Tag the exposed keywords in your sheet with a simple yes or no. You’ll use that flag in the next step.

3. Forecast Keywords With Semrush

Semrush is the fastest way to get per-keyword forecasts. Work through it in order:

1. Build the keyword set. Open the Keyword Magic Tool and pull the client’s targets, including long-tail and question keywords, not just the head terms.
2. Grab the data for each keyword. Take search volume from Keyword Overview, current position from Position Tracking, and estimated CTR by position from Organic Research. Note the Personal Keyword Difficulty and Potential Traffic figures too, since both weigh your own domain’s authority rather than a generic average.
3. Find the gaps. Run the Keyword Gap tool against two or three competitors to surface keywords they rank for and the client doesn’t.
4. Mark the AI Overview keywords. Semrush shows which keywords trigger an AI Overview. Flag those, because they need the discount in step five.
5. Do the math. For each keyword, Potential Traffic equals Search Volume times your AI-adjusted CTR, and Potential Conversions equals Potential Traffic times your conversion rate. Cut the CTR on the flagged keywords before you multiply.
6. Sanity-check the total. A keyword that jumps from zero to ten thousand visitors is a broken formula, not a discovery.

4. Forecast With SE Ranking’s SEO Potential Tool

SE Ranking does the CTR and traffic math for you, which is useful when you want a second number to check Semrush against:

1. Set up the project. Add the target keywords and current rankings to SE Ranking’s Rank Tracker.
2. Configure conversion settings. Set the conversion ratio and profit per conversion. If one in fifty visitors buys, that’s a 2% conversion rate.
3. Review current performance. Read the current traffic forecast, traffic cost, and estimated income based on where the site ranks today.
4. Model the ceiling. In the SEO Potential tool, set “Estimated top” to 1. That shows the maximum traffic and leads you could pull if you ranked first for every target keyword, which anchors the optimistic end of your range.
5. Read the outputs. The tool returns expected traffic volume, what that traffic would cost through Google Ads, and the expected number of prospects.
6. Compare competitors. Use the competitor tracking to benchmark your forecast against rivals’ actual performance.

Already on SE Ranking? Swydo pulls its Traffic Forecast straight into client reports, so the estimate lands next to the live rankings rather than in a separate export.

One honest limit: the SE Ranking integration doesn’t carry deep ranking history beyond average position, so the twelve months your statistical model needs still comes from your own GA4 and Search Console records.

5. Build the Model in Google Sheets

A spreadsheet gives you full control of every assumption and is the best format for a pitch, because you can show your working and model scenarios side by side. Build it once and reuse it per client.

Set up the columns. Create a sheet with these thirteen columns:

Column F, the AI Overview discount. This is the piece the old models miss. When an Overview shows, the ranking earns a fraction of its usual clicks, so apply a haircut:

=IF(C2="yes",0.4,1)

That cuts the click rate by 60% on any keyword with an Overview and leaves the rest at full value. Tune the 0.4 to your own data: lighter if your brand is often cited in Overviews, heavier if it never is.

Columns G and H, the click-through rates. Both turn a rank into a CTR off a position curve, then multiply by the discount. These are typical no-Overview values you can adjust to your own numbers. Current CTR reads the current rank in D:

=IF(D2<=1,0.28,IF(D2<=3,0.15,IF(D2<=5,0.09,IF(D2<=10,0.04,IF(D2<=20,0.015,0.005)))))*F2

Forecasted CTR is the same formula pointed at the forecasted rank in E:

=IF(E2<=1,0.28,IF(E2<=3,0.15,IF(E2<=5,0.09,IF(E2<=10,0.04,IF(E2<=20,0.015,0.005)))))*F2

Traffic and conversions. Current traffic is volume times current CTR (=B2*G2), forecasted traffic is volume times forecasted CTR (=B2*H2), and the two conversion columns multiply each traffic figure by your conversion rate.

Add summary stats below the table. Assuming your keywords fill rows 2 to 100, put this block underneath:

Chart it. Highlight the keyword column and the current and forecasted traffic columns, then Insert then Chart and pick a bar chart. Give the current and forecasted bars different colors so the uplift reads at a glance.

Google Sheets Forecast Example – Use this template and make a copy for your own forecasting model.

Build three scenarios. Copy the model onto three tabs named Best Case, Worst Case, and Most Likely, then change only the forecasted rank in column E on each. Best case assumes strong positions, worst case modest ones, most likely wherever your judgment lands. The forecasted CTR recalculates from the rank on every tab, so each one produces its own traffic and conversion total. The spread between the three is your forecast range.

Pressure-test it with a sensitivity table. On a tab named Sensitivity, list the CTR and conversion rate you expect at each rank band, using the rank floor as the lookup key:

Pull the conversion rate into the model with an approximate-match lookup on the forecasted rank:

=VLOOKUP(E2,'Sensitivity'!$A$2:$C$7,3,TRUE)

The rank floors have to be numbers, not text like “6-10”. An approximate VLOOKUP walks a sorted numeric column and grabs the last row at or below your rank, so a forecasted rank of 8 correctly lands on the row for 6. Text ranges break that match without an error and hand you the wrong rate, which is a common reason a spreadsheet forecast is quietly off. Nudge the numbers and watch the bottom line. If a half-point change in conversion rate swings the forecast hard, the client needs to know the estimate is fragile before they bank on it.

6. Project the Trend With Statistics

When an account has clean history, a statistical model reads the trend better than a keyword-by-keyword build. Use it alongside the keyword method, not instead of it.

1. Gather the data. Twelve or more months of organic traffic, rankings for key terms, and conversions, with major site changes and algorithm updates marked.
2. Spot the patterns. Look for seasonality and the overall direction before you model, so you know what the output should roughly look like.
3. Pick a model. Linear regression fits a straight-line trend and suits steady accounts. Exponential smoothing handles seasonal peaks and troughs better.
4. Run it. Excel does this with the FORECAST and TREND functions. For more control, a Colab notebook runs the same regression in Python: open the notebook, save a copy to your Drive, run each cell, enter your monthly data when prompted, and read the forecast and its range off the chart.
5. Interpret and refine. Add new data as it comes in and adjust for known future events like a planned migration. The forecast sharpens every month you feed it.

Without making you do all of the coding. I created a Python code using Linear Regression to create the forecast. Here’s how you can use it:

Open the Google Colab Link:

• Click the link to open the Colab notebook.

Make a Copy of the Notebook:

• Once the notebook is open, go to the top menu and click on File > Save a copy in Drive. This will create a copy of the notebook in their own Google Drive, allowing them to modify and run the code.
• Alternatively, they can click on File > Download .ipynb to download the notebook and upload it back into Colab later.

Run the Code:

• Once the copy is saved, they can run each code cell by clicking on the play button (or press Shift + Enter).
• When prompted, they can enter their monthly organic traffic data in the format Month,Traffic (e.g., 1,1000;2,1200;3,1300).
• The code will process the data and generate a forecast for the next 6 months.

View the Forecast:

• The notebook will display a chart showing both the historical data and forecasted traffic for the next 6 months.
• It will also print the forecasted traffic values, including a range for potential outcomes.

Pro Tip: Forecasts are guides, not guarantees. Use them to inform strategy, but be prepared to adapt as real-world data comes in.

7. Benchmark Against Competitors

When the client is a new domain with no history to project from, borrow it from the competition.

1. Find the real rivals. Look past who the client thinks they compete with. Focus on who actually ranks in the SERP for their target keywords.
2. Pull their performance. Estimate their organic traffic with a tool like SimilarWeb or Semrush, and note their rankings, content cadence, and backlink profile.
3. Find the gaps. Spot where the client lags and where they already have an edge, which separates the quick wins from the long game.
4. Set realistic benchmarks. Use a competitor on a similar growth path, not the category giant. The market leader’s traffic is a fantasy target for a new site, and setting it as the goal just sets up the disappointment.

How to Turn a Forecast Into Client Goals You Can Track

A forecast that lives in a spreadsheet is a forecast nobody checks. The value shows up when the projection becomes a tracked target the client can watch against real results, which is the difference between a pitch document and an ongoing reason to keep paying you.

Set the forecast as a goal you monitor. In Swydo, goal tracking turns each forecasted target into a pacing metric with On Track, Off Track, and Achieved states, so you and the client see at a glance whether the account is tracking toward the number you promised. The Recent Periods view shows the historical values while you set the target, so you’re benchmarking against what actually happened rather than a guess.

If you built your model in a spreadsheet, you don’t have to rebuild it. Swydo’s Google Sheets integration imports the forecast straight into the client report, where it sits alongside the live data from GA4, Search Console, and your rank tracker. The projected line and the actual line, in one place, is the single most persuasive thing you can put in a monthly SEO report for clients.

For the accounts that need it, SEO monitoring catches the moment reality diverges from the forecast, so a bad month becomes a conversation you start instead of one the client starts for you.

Is SEO Forecasting Still Worth It?

Yes, but not as a crystal ball, and anyone selling it that way is setting a trap. A forecast is a planning tool. It’s your best estimate given imperfect information, and its job is to make the decision about budget and effort a little less blind, not to guarantee an outcome no one controls.

The mistakes that wreck forecasts are consistent, and worth naming. There’s the straight-line assumption, when rankings really improve in lumps and plateaus. There’s the seasonality blind spot, where a normal Q1 dip reads as failure. Worst of all, there’s the stale click-through rate that predates AI Overviews, the big one now and the reason so many forecasts are already too optimistic. And then there’s the number-one ranking promised on a keyword the client has no realistic shot at, which feels good in the pitch and terrible in month six.

So what does your forecasting actually account for right now? If the answer is rankings and a fixed CTR, it’s measuring a version of search that’s already gone.

Final Thoughts

SEO forecasting still does the two jobs it always did: it wins the pitch and it sets expectations you can live with. What changed is the math underneath. Rankings no longer convert to clicks at the old rate, so a forecast that skips the AI Overview discount overpromises by design.

Build the baseline from clean twelve-month data. Flag the keywords exposed to AI Overviews and discount them. Pick a modeling method, hand the client a range, then track it against real results so the number stays honest. Do that and the forecast stops being a sales prop and becomes the thing that turns a new client into a long one.