Is SEO Worth It Now That Google Answers the Question Itself?

Is SEO worth it? That depends on what you were actually buying. Search still sends people who are ready to buy; what changed is that the traffic arriving before they are ready is thinning out. So the answer depends entirely on which of those two things you were paying for — and most people paying a retainer have never been told which one they were getting.

This page is a way to find out, including the version where you stop paying.

We should say our own position before we ask you to trust any of this: we sell marketing operations. We are not a neutral party. What we can offer instead of neutrality is our own measurements, our own method, and a section further down naming the conditions under which we would tell you to stop — because a page that cannot say that is an advertisement.

The Short Answer

Keep paying if your search traffic converts, you can see which searches produce enquiries, and the pages you are paying for answer things a buyer asks close to a decision.

Stop paying if you cannot name a single search term that produced a customer in the last six months, or your provider cannot either. That is not a hard-times judgement. It is a missing feedback loop, and no amount of further spend will produce one.

Change what you are paying for if the work is aimed at the top of the funnel — definitional articles, “what is” pages, volume — because that is precisely the traffic an AI answer is now absorbing.

What Actually Changed, and What Did Not

An AI Overview is a generated answer Google places above the organic results. Google’s own documentation is the place to understand what it is and how it is produced — Search Central describes AI Overviews as features that “help people get to the gist of a complicated topic or question more quickly, and provide a jumping off point to explore links to learn more.” It is not a place to learn whether search is still worth your money, and no vendor’s blog is either.

What changed is narrower than the headlines: for some queries, the answer that used to require a click is now on the results page. If your page’s job was to be that answer, its click-through is under pressure.

What did not change is the part that pays. Someone typing “emergency electrician” or “how much does a fractional CMO cost” is not looking for a paragraph. They are looking for someone to hire. Those queries still route to pages, and those are the ones with money attached.

This is why “is SEO dead” is the wrong question. It smuggles in the assumption that all search traffic is one thing.

What We Measured Ourselves

Here is the part we can offer that the pages competing with this one cannot: we run this playbook on our own site and we publish the readings.

Across eight of our own target queries, measured on Ahrefs US between 11 and 22 August 2026, an AI Overview was present at position 1 on six, and absent on two:

Our Own Overview Readings by Term and Date

What Our Own Overview Measurements Showed

Read that carefully, including its limits. Eight queries is a small sample, they are our target terms rather than a random draw, it is US-only, each is a single snapshot rather than a trend, and the readings are now around a month old. We are not claiming a law of search from eight rows.

But one thing in it is worth your attention: the two queries with no AI Overview were the two closest to a buying decision — “do I need an agency” and “is Google Ads worth it”. The ones carrying an Overview skewed definitional. That is a pattern in our data, not a proven rule, and the useful response is not to believe us. It is to measure your own terms the same way.

The method costs nothing to describe: list the ten searches you believe produce your customers, look at each results page, and note whether an AI answer sits above the results and whether it links anyone. Google reports a version of this from its own side too: Search Console now carries Search generative AI performance reports, which show how your own pages are doing inside Google’s AI experiences. You will learn more from that hour than from any article about the death of SEO, including this one.

The Question Underneath the Question: What Were You Buying?

Most people asking whether SEO is worth it are really asking whether a specific invoice is worth it. Those are different questions, and the second is answerable.

Ask your provider, or ask yourself if the work is in-house:

• Which searches were we trying to win, and why those? A list of terms with no stated reason is a list, not a strategy.
• Which of those searches produced an enquiry in the last six months? Not traffic. Enquiries.
• What did we publish, and what was each page’s job? “Ten blog posts” is a volume report. “Three pages answering the questions people phone us to ask” is a plan.
• What would you stop doing if the budget halved? The answer reveals what they actually believe is working.

If those questions produce confident, specific answers, you are probably buying something real. If they produce traffic charts and impressions, you are buying activity.

Impressions are the tell. They are the metric that survives every bad quarter.

Four Conditions That Decide It for You

These are the conditions we would actually check, in order. Any one of them failing does not settle it; the pattern does.

1. Attribution exists. You can tell, for a given enquiry, roughly where it came from. Not perfectly — nobody has perfect attribution — but you are not guessing. Without this, “worth it” is unanswerable by anyone, and the honest thing is to fix this first and defer the decision.

2. The work targets decision-stage searches. Look at what has been published in your name over the last six months and ask, of each page: would somebody near a purchase search for this? If most of the output is definitional, the strategy predates the change described above.

3. Somebody can name the win. A specific search term, a specific page, a specific customer. One is enough. If nobody — you, your agency, your in-house marketer — can produce one example after a year of investment, that is not a slow-burn problem.

4. The cost is in proportion to what a customer is worth. You need to know what one customer is worth to you before any of this is decidable. If you do not have that number, get it before you renew anything.

When the Honest Answer Is Stop Paying

This is the section a vendor cannot write, so here it is in concrete terms. Stop paying if any of these is true:

• Twelve months in, nobody can name a single search term that produced a customer. Not a ranking, not a traffic increase — a customer. Twelve months is past every legitimate “it takes time” window.
• Your reporting is impressions, rankings and traffic, with no line connecting to enquiries, and you have asked for that line more than once and not received it. The second clause matters. The first time is an oversight; the third time is an answer.
• The searches you are ranking for are ones no buyer of yours would type. Ranking first for something irrelevant is a cost, not an asset.
• Your enquiry volume from search has not moved in six months and neither has the work — the same volume of blog posts, the same monthly report shape, no change in approach despite a visible change in search itself.
• The retainer is large enough that you would notice it disappearing, and nobody is accountable for a number. That combination — real money, no owner — is the one that quietly runs for years.
• You cannot say what you are buying. If you have read this far and still cannot describe the work in a sentence, stop paying and ask again from a position of having stopped.

Stopping is not irreversible. Search work you have already paid for does not evaporate in a quarter — the pages stay published and keep doing whatever they were doing. That asymmetry is why stopping is a cheaper experiment than continuing, and it is exactly the experiment nobody selling you the service will propose.

When It Is a Vendor Problem Rather Than a Search Problem

Worth separating, because the fix is different and cutting the budget fixes the wrong one.

It is a vendor problem if: the reporting never changes shape, you are told rankings improved but cannot find the enquiries, the strategy has not been revisited since AI answers appeared, or every question about results is answered with a timeline.

It is a search problem if: your provider can show you the terms, the terms are genuinely decision-stage, the pages answer them well, and the enquiries still are not there. That happens. Some businesses’ buyers do not use search to find them, and a good provider will tell you so — the honest ones do, and it is the fastest way to identify one.

The difference is diagnosable in an afternoon with the four conditions above. Do not skip to cancelling the invoice; you may be cancelling the wrong thing.

Where We Would Fail This Test

Applying our own conditions to ourselves, in public, because a page that only points outward is not worth much.

We would fail condition 3 today. This site is young and starting from a Domain Rating of zero — we published that measurement, and what it costs us, on our content marketing page. We cannot currently point at a search term that produced a customer for us, because there has not been enough time for one to exist. By the standard we just set out, a reader would be right to discount anything we implied about our own results — which is why this page rests on measurements and method rather than on outcomes we have not earned.

We are also not neutral. We sell marketing operations. If you conclude from this page that your provider is fine and you should leave things alone, that is a worse outcome for us than if you conclude you need help. We have tried to write it so that the first conclusion is genuinely reachable, and the stop-paying list above is the test of whether we managed it.

How to Decide This Quarter

1. Get the customer-value number. Average sale × repeat × margin. Rough is fine; it just has to exist.
2. List the ten searches you think produce your customers. Then check each one yourself — is there an AI answer above the results, and does it link anyone?
3. Ask the four questions in “what were you buying?” Write down the answers rather than accepting them conversationally.
4. Run the stop-paying list. If two or more are true, stop, and use the pause to find out whether anything changes.
5. If you keep paying, change the brief: fewer definitional pages, more pages answering what people ask you on the phone at the point of deciding.
6. Set the review date now, and the number you will review against, before the next invoice.

If step 1 or step 2 stops you, that is the page working. Both are free, and both are yours to keep whatever you decide about the invoice.