How Much Traffic Should You Expect From SEO?
A traffic forecast is not a promise, it is a claim about clicks that already exist and positions a competitor currently holds. That makes it auditable. Most projections handed to buyers fail one test: the clicks they forecast exceed the clicks the category has.
A Traffic Forecast Is a Claim, Not a Promise
A traffic forecast is a claim about clicks that already exist and positions a competitor already holds. That is what makes it auditable, and almost nobody audits it.
Before a twelve-month SEO budget gets signed, someone produces a number. It usually arrives as a single confident line on a slide: organic sessions climbing month over month to some figure a year out. That number is the most consequential and least examined artifact in the whole engagement, because the spend is justified by it while nobody checks the arithmetic behind it.
It can be checked. Both halves of the claim are observable facts about the world today. The first half is how many clicks a given set of queries actually produces, which is bounded: Google states it sees more than 5 trillion searches annually, and any category is a small, countable slice of that. The second half is which competitors currently occupy the positions that earn those clicks, which anyone can look up in an afternoon. A forecast that cannot be traced back to those two facts is not a projection. It is a wish with a chart attached.
| Input | What the slide assumes | What the evidence supports |
|---|---|---|
| Position reached | Best case, usually first | A band set by the authority gap |
| Click rate by position | A vendor benchmark table | Roughly 16 percent odds lost per position |
| Clicks available | All of the search demand | Deflated to 8 percent under an AI answer |
| Time to arrive | A smooth ramp from month one | Hours to several months, per Google |
| Output unit | Sessions | Pipeline, with a confidence band |
| Validation | None | Total cannot exceed the clicks that exist |
Each row on the right is a correction that pushes the number down, which is why honest forecasts look smaller than the ones that win pitches. The five sections that follow build the number properly, one stage at a time, then the last one shows how to take apart any projection already sitting on your desk.
Start From the Demand That Exists, Not the Keywords in the Deck
The first stage of a defensible forecast is a ledger of individual queries, each with its own monthly demand and its own commercial intent. Not a category, not a theme, not a total. Line items, because every correction that follows has to be applied per query, and an aggregate number cannot carry them.
This is where most decks quietly break. They are built on a short list of fat head terms, which is the demand a stranger would guess a business wants, rather than the demand that exists. Google is explicit that its query stream does not work that way: in its own How Search Works documentation it states that 15 percent of the searches it processes every day are ones it has never seen before. A meaningful share of real demand is not on any keyword list, because it has never been typed before today.
That tail is getting longer, not shorter. Google reports that the average AI Mode search in the United States is now triple the length of a traditional query. Longer queries are more specific queries, and more specific queries spread the same underlying demand across many more distinct strings. A forecast built on twenty head terms is describing a smaller and smaller portion of the market it claims to address.
Two facts bound the whole exercise before any multiplication starts. Almost all of this demand sits on one engine, which holds 91.27 percent of the global search market, so the forecast is effectively a claim about one company's result pages. And a large share of those searches never send anyone anywhere: Pew Research found that around two-thirds of searches ended without a click to any external site at all. The pool is one engine wide, and most of it is already closed.
A good query ledger therefore looks unglamorous. It is long, it is weighted toward specific phrases with modest individual volumes, and it carries an intent label on every row so that the conversion step later has something real to work with. If the forecast you were handed fits on one slide, it has not been built yet.
The Position You Can Reach Is Set by the Authority Gap
The second stage assigns each query a position band rather than a position. The band is set by the distance between your site's authority on that topic and the authority of the specific competitors who hold those slots today, because ranking is not a score you achieve in isolation. It is a place you take from somebody who currently occupies it.
Every forecast that opens with an assumed first position has skipped this stage. Google's own guidance for buyers is unusually blunt about it, and it is worth reading as a procurement instruction rather than as advice.
No one can guarantee a #1 ranking on Google.— Google Search Central, guidance on hiring an SEO
The same Google page goes further, telling site owners that a provider who guarantees first place is a reason to find someone else. A forecast is not a guarantee, so it does not fall foul of that directly. But a forecast whose arithmetic silently assumes first place on every target query has smuggled the guarantee in through the spreadsheet, and it should be treated with the same suspicion.
The band also has to be spread over time, and here too the platform contradicts the usual slide. Google's SEO starter guide states that some changes might take effect in a few hours while others could take several months, and advises waiting a few weeks before judging whether work had any beneficial effect. That is the platform describing its own latency. A curve that ramps smoothly from month one is not modelling that latency; it is modelling a spreadsheet formula. The realistic shape is flat-to-negligible early while indexing and authority accumulate, then compounding, which is also why the timeline question deserves its own honest answer rather than a hopeful gradient.
Practically, this stage produces three numbers per query instead of one: a conservative position, an expected position, then an optimistic one, each justified by naming who currently holds it and what would have to change for them to be displaced. If a forecast cannot name the incumbent it plans to overtake, it has not made a claim about the world. It has made a claim about ambition.
Demand Is Not Clicks: The Two Deflators Every Forecast Omits
The third stage is where the largest errors live, because two separate reductions sit between a query's search volume and the clicks a site can actually earn from it. Most projections apply neither, which alone is enough to inflate a number by a multiple.
The first deflator is position decay, and it is better grounded than the vendor tables usually cited for it. A set of three controlled experiments with 3,196 participants, published in a peer-reviewed journal, measured the effect directly: every one-position drop in rank cut the odds of a result being clicked by about 16 percent, an odds ratio of 0.84 that held with result accuracy and search engine controlled for. What matters for a forecast is the shape rather than the exact coefficient. The decay is multiplicative and constant per position, not a cliff after the top three, so position five is worth roughly half of position one, then position ten roughly a fifth.
Read that curve against the position band from the previous stage and the effect on a forecast is immediate. Moving an expected position from first to fourth does not shave a little off the projection. It roughly halves it, on every query where the assumption was made.
The second deflator is the result page itself. Pew Research measured actual browsing behaviour. When an AI Overview style summary was present, users clicked a traditional result in 8 percent of visits against 15 percent when no summary appeared, then clicked a link inside the summary itself just 1 percent of the time. The clickable share of a query nearly halves when the answer arrives above the results, and these summaries are not a fringe surface: Google reports AI Overviews reaching more than 2.5 billion users a month.
Honesty requires the counterweight, because the pessimistic reading of those numbers is also wrong. Organic listings have not stopped receiving clicks. An eye-tracking and mouse-tracking study of 2,776 commercial-intent result pages found that 82.42 percent of clicks landed on non-ad elements, with advertising in all its forms taking the rest. The sample is small at 47 participants, so treat it as directional. Its direction is clear enough: on the queries where money changes hands, the large majority of clicks still go to unpaid listings. The pool is smaller than the deck assumes, then it is still overwhelmingly organic.
Want the Forecast on Your Desk Audited Before You Sign It? Digital Strategy Force rebuilds it query by query with both deflators applied, then tells you which number survives.
Convert to Pipeline, Because Traffic Is Not the Deliverable
The fourth stage turns surviving sessions into pipeline using intent-weighted conversion and deal value. This is not a presentational flourish. A session count is close to unfalsifiable in a board meeting, and it invites the wrong optimisation, because the cheapest way to raise a session number is to chase queries that never buy anything.
Conversion also cannot be a single site-wide average applied to every row. The intent label attached in the query ledger is what makes this stage meaningful: a comparison query, a specification query, then a brand query all convert at different rates into different deal sizes, so each block of the ledger carries its own rate. Averaging them together produces a number that is arithmetically fine and commercially meaningless.
There is a reason this stage now matters more than it used to. Microsoft's Bing team reports that AI search has not stopped clicks so much as moved them later in the journey with far stronger intent, noting that Copilot-assisted journeys run 33 percent shorter while high-intent conversion rates are 76 percent higher than on traditional surfaces. If that holds even partially, a forecast measured in sessions will understate a good program and overstate a bad one, because the visits it counts are no longer interchangeable.
The buyer-side shift points the same way. Bain & Company found 44 percent of United States online buyers now begin product research in an AI tool or split it between AI and traditional search, while Stanford's AI Index recorded organisational generative-AI use rising from 33 percent to 71 percent in a single year. Fewer, later, better-qualified visits is the shape of the channel now, which is precisely why pipeline is the only honest output unit.
Notice what that table implies about incentives. Every stage that improves a forecast's honesty also shrinks it, so a vendor competing on the size of its projection is structurally rewarded for skipping stages. That is the whole reason two agencies quoting the same query set can hand you numbers a factor of three apart.
The DSF Organic Forecast Pipeline
Organic projections at Digital Strategy Force run through five ordered stages, the DSF Organic Forecast Pipeline, and the order matters because each stage consumes the output of the one before it. The governing rule is a single sentence: a forecast is a claim about clicks that exist and positions someone else currently holds.
Stage one is the Query Ledger, which enumerates real queries as line items with demand and intent, accepting that a meaningful share of demand sits in the long tail rather than in head terms. Stage two is the Authority Gap, which converts each row's ambition into a position band justified against the named incumbents holding those slots. Stage three is Click Yield, which applies both deflators, per-position decay then result-page deflation, to turn positions into clicks. Stage four is Pipeline Conversion, which reprices surviving sessions as pipeline at each row's own conversion rate and deal value.
Stage five is the Ceiling Check, and it is the one that does the real work in a sales conversation. It applies the Addressable-Click Ceiling: the total forecast can never exceed the clicks the category actually produces. Google's disclosure of more than 5 trillion annual searches is the outer bound of the entire system, and any single category is a countable fraction of it. Sum a projection's forecast clicks, compare against the category's addressable pool, and a projection that breaches the ceiling is not optimistic. It is arithmetically impossible, and it is rejected rather than negotiated.
| Stage | What it computes | What a slide does instead | Source of truth |
|---|---|---|---|
| Query Ledger | Demand per query, with intent attached | Twenty head terms | Query-level demand |
| Authority Gap | A position band against named incumbents | Assumes first place | Current result pages |
| Click Yield | Clicks after position decay and page deflation | Applies neither | Published research |
| Pipeline Conversion | Pipeline value per row, not sessions | Reports sessions | Your own close data |
| Ceiling Check | Whether the total is even possible | No validation step | Addressable clicks |
The pipeline is deliberately unkind to its own output. Run honestly it produces a smaller number than the one that won the pitch, expressed as a band rather than a point, denominated in pipeline rather than sessions. That is what makes it defensible twelve months later, when someone asks whether the program did what was promised.
How to Stress-Test the Forecast on Your Desk
You do not need to rebuild a vendor's model to judge it. Three structural errors account for nearly every projection that misses by a factor of three, and each one leaves a visible fingerprint on the document you were given.
The first is the best-case position assumption. If the model does not state a position band per query, and does not name who currently holds those positions, it has assumed the best case. The second is the missing deflators. If the words describing position decay and result-page deflation do not appear anywhere with numbers attached, the clicks have been overcounted twice over. The third is unreachable demand, where the forecast is built on head terms the site has no authority to compete for, which is the error the Authority Gap stage exists to catch.
| Error | The fingerprint in the document | Which stage catches it |
|---|---|---|
| Best-case position | No position band, no incumbent named | Authority Gap |
| Missing deflators | Neither position decay nor AI deflation appears | Click Yield |
| Unreachable demand | Built on head terms far above the site's authority | Query Ledger, then Ceiling Check |
Then run the ceiling test, which takes about ten minutes and settles most arguments. Add up the forecast's projected clicks for the year and compare that total against the plausible click pool for the category, which is the sum of the demand across the queries in scope, discounted by the deflators. If the projection needs more clicks than the category produces, no amount of execution can deliver it. This is also the test that exposes a subtler failure: forecasts that quietly assume the business will win a majority of every query it targets, simultaneously, against incumbents who are not standing still.
| Ask of the document | Passes if | Red flag if |
|---|---|---|
| Is the output a range? | ✓ A band with scenarios | ✗ One confident line |
| Are the incumbents named per query? | ✓ Yes, with the gap stated | ✗ Positions asserted |
| Is position decay applied? | ✓ Stated as a number | ✗ Not mentioned |
| Is AI-answer deflation applied? | ✓ Stated per query class | ✗ Not mentioned |
| Is the unit pipeline? | ✓ Revenue or pipeline | ✗ Sessions only |
| Does it pass the ceiling test? | ✓ Total fits the pool | ✗ Total exceeds the pool |
There is a version of this conversation that goes badly, and it is worth naming. A buyer who rewards the largest forecast trains every vendor to produce one, and the vendor who arrives with the honest, smaller, banded number loses the pitch to the one who skipped four stages. The only defence is to judge the method rather than the magnitude, which is precisely what the scorecard above is for.
It is also worth being clear about what a good forecast is not. It is not a promise, and it is not a substitute for judgment about whether the channel suits the business at all. Organic search remains an owned position that keeps returning after the spend, which is a different asset class from renting clicks, and the paid alternative has its own trajectory: Alphabet's filings show cost-per-click rising 7 percent while paid click volume grew 5 percent.
Neither channel is free of gravity. The difference is that one of them can be forecast, audited, then held to account, and now you know how to do it. If the number you were handed cannot survive the six questions above, the problem is not that the forecast is optimistic. The problem is that it was never a forecast, and the spend it justifies deserves a better basis than that.
FAQ — SEO Traffic Forecasts
How much traffic should you expect from SEO in the first year?
There is no benchmark answer, and any figure quoted without seeing your query set is guesswork. The honest output is a band built query by query: real monthly demand, an achievable position range set by the gap to the incumbents holding those positions, then both click deflators applied. Google states changes take from hours to several months to register, so a first-year curve should start near zero and compound rather than open at its target.
Why do SEO traffic forecasts vary so much between agencies?
Because three assumptions usually go unstated, and each one multiplies. One forecast assumes first position while another uses a realistic band. One applies a full click-through rate while another deflates for the result page. One targets head terms the site has no authority to reach. Change those three and the same query set produces answers a factor of three apart, which is why the assumptions matter far more than the headline number.
Can an agency guarantee a number one ranking or a traffic number?
No. Google says so in its own guidance for hiring an SEO: no one can guarantee a top ranking, and a provider promising first place is a reason to walk away. A number can be forecast with stated assumptions and a confidence band. It cannot be promised, because the positions in question are held by competitors who also get to act.
How long before SEO traffic actually shows up?
Google's starter guide states some changes take effect in a few hours while others take several months, and advises waiting a few weeks before assessing whether work had a beneficial effect. In practice the early months buy indexing and authority rather than sessions, so a forecast showing meaningful traffic in month one is describing something other than organic search.
Should a forecast be expressed in traffic or in revenue?
Pipeline, with sessions shown as an intermediate step. A session count is close to unfalsifiable in a board meeting, then it invites the wrong optimisation, because the cheapest way to raise sessions is to chase queries that never buy. Converting through intent-weighted rates also captures something traffic hides: clicks now arrive later in the journey with stronger intent, so fewer sessions can be worth more than the ones they replaced.
Do AI Overviews make SEO traffic forecasts worthless?
They make the old arithmetic wrong, not the exercise pointless. When an AI summary appears the result-click rate roughly halves, so any forecast ignoring it overstates the outcome. On commercial queries, though, the large majority of result-page clicks still land on unpaid listings, so the correct response is to deflate the forecast and shift the goal toward being the cited source, never to abandon forecasting altogether.
Next Steps — SEO Traffic Forecasts
- ▶ Ask for the query ledger rather than the total. Any forecast that cannot be decomposed into named queries with their own demand and their own incumbents is a guess wearing a chart.
- ▶ Make both deflators explicit. Require the position-decay assumption and the AI-answer deflation to appear as stated numbers; if they are absent, the projection is overstated by construction.
- ▶ Run the ceiling test yourself. Sum the forecast clicks, compare against the category's addressable pool, then reject anything that needs more clicks than the category produces.
- ▶ Convert it to pipeline before approving it, so the number competes honestly against every other line in the budget instead of hiding behind sessions.
- ▶ Check whether the traffic you already lost is recoverable, because organic traffic absorbed by AI answers changes the baseline any forecast is measured against.
Digital Strategy Force builds organic forecasts bottom-up through all five stages, then stress-tests whatever projection is already on your desk. Talk to the SEO team.
Open this article inside an AI assistant — pre-loaded with DSF's framework as the lens.