SEO for lead generation: what search delivers and where the leads are lost

The number that opens most pages written for this query is a close rate: SEO leads close at 14.6 %, outbound leads at 1.7 %. Everyone repeating it credits Search Engine Journal, and none of the pages carrying it names a sample, a window, or what counted as a lead. What makes it worth quoting is not that it travels without a source. It is that a close rate describes a sales process, and it is being used to sell a channel whose work ends at the click.

I sell this work, which is the reason to be exact about what it does. Search can put a page in front of somebody who typed what you sell. It cannot answer the phone.

Does SEO generate leads?

SEO generates visits from people already searching; what turns a visit into a lead is the page, the form and the follow-up behind it. A local business converts on a phone number and a map; a professional services firm on a specific page and a fast reply; a long-cycle B2B buyer converts weeks later, off-site.

Quick answer

• SEO produces visits. The lead is produced by what sits behind the page. The form, the offer and whoever replies decide it, and a search engine touches none of the three.
• Query intent decides the lead more than position does. Ranking first on a definitional search fills a traffic report and leaves the pipeline where it was.
• Attribution miscounts in both directions. Google Analytics 4 documents that all of its models exclude direct visits from receiving credit, which is where long decisions end up.
• The channel has preconditions. Without searchable demand, with revenue needed this quarter, or with no capacity to serve more enquiries, the money buys a worse problem.

What kind of problem this usually is

The model that wastes the most money treats SEO as a lead source, sitting in a list next to a referral partner and a trade show. Referrals arrive as people. Search arrives as sessions, and the distance between a session and a person is where the entire disappointment lives.

It shows up in the shape of the enquiry. A business asks for SEO because the phone is quiet, and the quiet turns out to have a nearer cause: a contact form that fails silently on mobile, a number that rings in an office nobody staffed after four, a reply that goes out two days later to somebody who called three suppliers on Tuesday. None of those is fixed by ranking. All of them are fixed cheaper than a six-month engagement, which is the uncomfortable part for whoever sells the engagement.

The mechanics of the channel itself, and what it can fairly be held responsible for, sit in what SEO is and what it is for. What follows here is the other half: the part of the path that search hands over to the business, and the places it comes apart.

The chain from impression to lead, and where it snaps

There are six links, and only the first three belong to a search engine.

1. The query exists and somebody types it.
2. The page is stored and shown for that query.
3. The click happens, which position alone does not guarantee.
4. The page answers what was typed, rather than what you wanted to say.
5. The visitor takes an action that leaves you a way to reply.
6. Somebody replies, in a window where the answer is still wanted.

Link three breaks more often than the reports suggest. In an audit I ran on a 312-page recipe site, one page held position 1 for its query with 124 impressions and zero clicks, because the title Google displayed did not contain the dish people were typing. The position was perfect and the traffic was nil, and no amount of new content would have moved it.

The scale of that gap is visible in two sites I own, one growing and one started from nothing, consulted through the Search Console API on 2026-08-13:

Fifty-seven clicks in a month is not a figure to plan a pipeline around, and reading the impressions beside it as an audience would misstate the situation by three orders of magnitude. The impressions column is the one that ends up in the report, and it is the one furthest from a customer.

The column that actually tracks the chain is the last one. Between those two months the pages printing impressions barely tripled while the pages reaching a human multiplied by ten, which is a different event from growth in visibility and the only one of the two with a person on the other end. That ratio is the closest thing this channel has to an early lead indicator, and almost no report shows it.

Intent decides the lead; position decides the traffic

The most expensive misallocation in this work is ranking beautifully for a question nobody asks before buying. It costs twice: the work is spent, and the traffic curve it produces makes everyone reluctant to admit the pipeline never moved.

The clearest signal of what a query is worth costs nothing to read, and it comes from the advertising auction. Ahrefs prices a click on this article's own query at 9 USD in the United States, over 1,800 monthly searches at a difficulty of 8, in a pull made for this blog on 17 August 2026. Nobody pays nine dollars for a definition. Where the paid market is quiet the query is usually informational, and the trade between renting those clicks and earning them is worked through in SEO vs Google Ads.

There is a second reason intent now outweighs position, and it is recent. Seer Interactive measured 3,119 informational queries across 42 organizations and 25.1 million organic impressions between June 2024 and September 2025, reporting a 61 % decline in click-through where an AI Overview appears and a 41 % decline where none does. Part of the fall has nothing to do with the summaries. What the two figures describe together is an informational half of the web losing clicks while commercial queries hold up far better, so the pages that were already weakest at producing enquiries are also the ones losing the visits.

Zero-click searching frames the same point from the other side: Similarweb puts it at 69 % of searches by May 2025, up from 56 % a year earlier. Selecting queries by volume in that environment selects for the traffic that never leaves Google. The free ways to find the alternatives are collected in keyword research without paid tools.

The same rule, at three scales

A local business wins on proximity queries with a decision attached, where the searcher is choosing this afternoon and the work is mostly profile, reviews and hours, which is the terrain of local SEO for small businesses.

A professional services firm competes on a handful of problem-shaped queries per specialty. The volumes are small, the value per enquiry is not, and the page that wins is usually the one describing the situation rather than the service.

A B2B buyer on a one-year cycle searches a few times, months apart, and converts off-site after a committee has read. The pages have to be there each time and none of those visits will look like a lead, which is examined in SEO for B2B.

What happens after the click, which is not SEO

A visit arrives with a question already formed. The page either answers it in the first screen or spends the visit explaining the company.

The failures at this stage are dull and they repeat everywhere. A form that demands a company size and a budget range before it will accept a name, on a first contact where the visitor has no reason to hand over either. A phone number rendered as an image on mobile. A thank-you page promising a reply within 48 hours to somebody comparing three suppliers this morning. None of these appears in a rankings report, all of them appear in the pipeline, and the ones I find most often are the ones nobody has opened on a phone since the site launched.

The cost comparison is what makes this worth saying out loud. SE Ranking surveyed 260 agencies across North America, Europe and Oceania, published on 13 December 2024, with its authors warning that the sample may not reach statistical significance; in it, 64 % of agencies charge under 1,000 USD a month. A year of that spend measured against an afternoon spent fixing a form is not a close comparison, and what each band actually buys is set out in what SEO costs a small business. My own diagnosis is priced at 500 USD for exactly this reason: whether the problem sits upstream or downstream of the click is worth knowing before anyone commits to months of work.

Attribution miscounts in both directions

Both errors are documented behaviour rather than bad luck, and they run in opposite directions.

Search gets undercredited. Google Analytics 4 states that all attribution models exclude direct visits from receiving credit, unless the whole path consists of direct visits, and that its data-driven model can reassign a conversion for up to seven days after it happened. A prospect who read a guide in March, remembered the name in November and typed the domain arrives as direct traffic, and the page that started the relationship is credited with nothing.

Search gets overcredited too, and this half is rarely mentioned by the people selling it. A buyer referred by a colleague, who searched the brand to check it existed and then called, appears in the report as organic search. The channel collects the credit for a decision made in a conversation it never touched. Both errors survive because the same report is used to defend the budget and to judge the work, which are two jobs with opposite incentives.

What survives contact with a long cycle is humbler than a dashboard. Which queries print impressions, which pages get entered from search, and whether callers start mentioning pages that were never advertised. The setup that makes the first two visible is covered in how to set up Search Console, together with a limitation worth stating plainly: that report knows queries, clicks and positions, and knows nothing whatsoever about revenue, so joining the two halves stays a manual job somebody has to decide to do.

The third measurement is the one nobody automates. A line on the intake form asking how the caller found you produces a messier answer than any attribution model and a more honest one, because it is the only source that can say the person read something in March and rang in November. When the curve moves the wrong way, the order of diagnosis then matters more than the explanation, which is the subject of diagnosing an organic traffic drop.

What the business has to be able to do

The channel makes a demand of the business that never appears in the proposal: it has to be able to absorb what it brings. A plumber with a full week turns a fifth enquiry into a missed call rather than revenue, and the site that produced it gets blamed for a capacity problem.

When SEO is not the channel

When the demand does not exist. If nobody types queries related to what you sell, there is nothing to capture, because search captures demand rather than creating it. A properly seeded harvest that comes back empty is real information: two of the twelve English queries measured for this blog's own plan returned zero relevant suggestions once they were actually seeded.

When the cash cycle is short. A channel measured in months cannot pay a bill due this quarter, and forcing it produces the timeline promises that characterise people who take money and disappear.

When the business cannot serve more enquiries. Capacity is a precondition, not a detail. Buying visits for a service already booked out converts interest into bad reviews.

When the sale does not begin with a search. Tenders, fixed vendor lists and referral-driven practices are reached another way, and ranking does not contest a list you are not on.

I turn down work on those four grounds more often than on price, and the reason is selfish as much as ethical: an engagement that cannot produce a result produces a client who says so. The questions that separate a serious proposal from a template are in how to choose an SEO agency.

What the first page of results leaves out

The prediction going in was that these pages would promise leads without ever distinguishing a visit from a lead. Reading them on 5 September 2026, that is half right, and the half that is wrong turns out to be more interesting.

The better pages do draw the distinction. Salesforce separates traffic generation from lead generation explicitly and covers forms, lead magnets and organic conversion rate. Nutshell says plainly that attracting visitors is only the beginning of the process. What none of the three most cited results contains is a section on when this channel is the wrong one, and none of them treats attribution as a source of error rather than a reporting feature. The advice is built so that every road ends in more SEO, which is what advice written by the seller tends to look like.

The numbers on those pages follow the same pattern as the close rate at the top of this article. One of them asks the reader to accept that 59 % of B2B and 49 % of B2C marketers trust this channel, sourced to a chart from 2012 and reprinted in 2026 without a window. Another opens by claiming that people process visuals 60,000 times faster than text, which arrives with no source at all and never had one. The pages that do attribute, Salesforce among them, name a report rather than a sample and a period. None of that is quoted here as evidence, and the figures in this article that carry a card are named with theirs.

Mistakes that repeat

• Hiring SEO to fix a conversion problem. The cheapest diagnosis in this work, and the one most often skipped, is checking whether the visits already arrive.
• Choosing queries by search volume. That filter selects for searches with no decision attached and discards the ones with a budget behind them.
• Judging the channel on last-click conversions. The model discards direct visits by design, and long decisions come back as direct.
• Counting form submissions as leads. A contact nobody qualifies is a row in a database, and the ratio between rows and conversations is where the real number lives.
• Reporting impressions while clicks stay flat. Two of my own sites show the gap: 102,472 impressions against 697 clicks in one month, 15,224 against 57 in another.
• Replying slowly. The enquiry that search produced expires on a schedule set by whichever competitor answered first.

Whether the agency doing this work is producing anything is a separate question with its own tests, and they are set out in is my SEO agency working.

Data and transparency

The 14.6 % and 1.7 % close rates in the opening, the 59 % and 49 % marketer-trust figures dated 2012 and the claim about processing visuals 60,000 times faster than text are quoted from the first page of results for this query to be examined, not used: none of the pages carrying them names a sample, a window or a method, and none traced to a primary measurement when checked on 5 September 2026. The three pages read for that section were Nutshell, Salesforce and Salesmate, which are among the first results for this query on that date. The click-through figures come from Seer Interactive's study published 4 November 2025, covering 3,119 informational queries across 42 organizations and 25.1 million organic impressions, measured from June 2024 to September 2025, which reports a 61 % decline where an AI Overview shows and a 41 % decline where none does. The zero-click figures, 56 % rising to 69 % between May 2024 and May 2025, come from Similarweb's Generative AI Landscape 2026 report. Agency pricing comes from SE Ranking's survey published 13 December 2024 over 260 responses from North America, Europe and Oceania, cited together with its authors' warning that the sample may not reach statistical significance. The attribution behaviour is Google's own documentation for Analytics 4, opened on 5 September 2026. The search volume, difficulty and cost per click for this article's query come from an Ahrefs pull for this blog on 17 August 2026, US market.

The monthly figures belong to westchesterride.com and voytara.com, two sites I own, so they are named with exact numbers, consulted through the Search Console API on 2026-08-13. Those impression counts sit across a known defect: Google's Search Console data anomalies page records a logging error that prevented impressions from being recorded accurately from 13 May 2025 to 27 April 2026, affecting impressions, click-through rate and average position but not clicks. The argument here rests on the clicks column, which is the one Google declares intact. The position-1 page with 124 impressions and zero clicks is client work, described by type and not named. My own prices are published on this site. No conversion rate, lead volume or timeline is given for any scenario here, because I have no measurement that would support one, and the point of the article is that the ones in circulation do not either. Verified as of August 2026.

Primary sources, opened on 5 September 2026: the Seer Interactive study; the SE Ranking survey; Google's attribution models documentation; the Search Console data anomalies page.

What this changes

The request that arrives is almost always for more traffic, and the honest first move is usually to argue against it.

A site with visits and no enquiries has a conversion problem that more visits make more expensive rather than less. A site with enquiries and no capacity has a business problem that ranking would convert into bad reviews. Both are common enough that the first hour of any engagement I take goes to finding out which one is on the table, and often enough the answer means the work I would have sold is not the work that was needed. Losing that sale is cheaper than the alternative, which is six months of correct SEO landing on a page that was never going to convert, and an invoice that has to explain itself.

Frequently asked questions