When it's worth it, and when it isn't
The people who say SEO was a waste of money are usually right about their own case and wrong about the general one. Their engagement failed for a specific, findable reason. Here's the short version before the long one.
| Your situation | Worth it? | Why |
|---|---|---|
| People already search for the problem you solve | Usually the best return available to you | Capturing demand is far cheaper than creating it, and it compounds |
| You sell locally and buyers search "near me" | Yes — and fast, by SEO standards | The map pack moves in weeks and most local competitors have neglected it |
| You invented the category | Not yet | Nobody searches for a thing they can't name. Create demand first, capture it later |
| Your site sits behind someone else's ticket queue | No, until that changes | SEO is analysis plus shipping. Without shipping it's an unread document |
| You need customers this quarter | No — run ads instead | Three to six months is the honest minimum before meaningful movement |
| Order value under ₹500, no repeat purchase | Rarely | The arithmetic needs traffic volumes most small sites won't reach |
Failure one: nobody is searching for what you sell
This is the failure nobody catches early, because keyword tools will always hand you a list. The list just doesn't contain enough of the queries that come from a person about to spend money.
You can test it before you spend a rupee. Write down the ten phrases someone would type in the week they're ready to buy from you — not the research phrases, the buying ones. Pull their monthly volumes. Then do the arithmetic honestly.
- Sum the monthly searches across those ten commercial phrases. Call it 300.
- Assume you get to position two eventually and take a generous share of the clicks — somewhere around 20–30% on a clean result page, and materially less when an AI answer sits above it. Call it 25%, so 75 visits.
- Apply your site's real conversion rate, from your own analytics, not a benchmark. For most B2B sites it's 1–3%. Call it 3%.
- You're looking at roughly two leads a month at full maturity, twelve months out, for a ₹75,000 retainer.
Failure two: a site nobody is allowed to change
This one is the most frustrating because everybody involved is doing their job. The agency writes recommendations. Your development partner has a six-week queue and a fixed-scope contract. Legal reviews the copy. Global HQ owns the CMS template and doesn't take requests from one market. Nine months later, 40 recommendations are sitting in a spreadsheet with two of them shipped.
Nobody decided this. It's the accumulated result of arrangements that each made sense on their own, and it's one of the most common ways an engagement quietly dies. The agency looks incompetent and the client feels cheated, and neither is true.
The test is simple: can a title tag change go live this week without a meeting? If the answer is no, fix that before you sign anything. A standing weekly dev slot, editable page content, and one person authorised to approve copy will do more for your results than doubling the retainer.
Failure three: the leads arrive and the sales process drops them
Organic leads are colder than a referral and warmer than a cold call, and they arrive at 11pm on a Sunday because that's when people research. If nobody responds until Tuesday, they've already spoken to two competitors.
The tell is specific and it's worth checking before you blame the channel: organic form fills went up, demo requests went up, and the founder still says "nothing's happening". Both statements can be true simultaneously, and the problem isn't upstream.
- The form goes to an inbox nobody owns. Marketing set it up, sales never agreed to watch it, and it fills quietly for six months.
- No CRM, so no attribution. Deals close and nobody can say where they came from. Organic gets credited with zero because zero is the default.
- "The leads are bad" with no definition of good. Ask sales to write down what a qualifying lead looks like. If they can't, the disqualification is a feeling.
- Response time measured in days. For high-intent organic enquiries, the first credible response usually wins. This is the cheapest thing on the list to fix.
Failure four: quitting at month four
Month four is where the cost curve and the return curve are furthest apart. The audit is paid for. The technical fixes are done and invisible. The content is published and hasn't ranked yet, because new pages take time to earn trust. Links haven't compounded. And the invoice has landed four times.
It is the rational-feeling moment to stop and the worst possible one, because everything you've paid for is about to start returning. The compounding in SEO is real and it's back-loaded — months seven through twelve typically deliver several times what months one through six did, on the same monthly spend.
The defence against this is not faith. It's leading indicators. Impressions for your target queries should be moving by week six. Average position on the specific keywords in the plan should be improving by month three. Indexed pages should track published pages. If those three are flat at month four, quitting is correct and you should. If they're moving and revenue isn't yet, you're on schedule. We wrote about why clients fire their SEO agency in month four because it's the most expensive mistake in the category.
The precondition that beats budget
If you gave us one variable to predict whether an SEO engagement will work, we wouldn't pick the retainer size, the industry, or the state of the site. We'd pick this: is there one named person inside the company who owns SEO and can get things shipped?
₹40,000 a month with that person in place beats ₹1,50,000 a month without them, and it isn't close. The reason is mechanical. SEO produces a continuous stream of small decisions that need someone with context and authority — which product page is the real one, whether this claim can go on the site, whether the dev team can move a template this sprint. Route those through a monthly call and each one costs three weeks.
- They don't need to know SEO. They need to know your business and be allowed to say yes.
- Two to four hours a week is enough. Less than that and decisions queue up.
- They approve copy, unblock dev, and answer "which of these three pages should rank" without escalating.
- They hold the baseline number and report it internally, so the engagement is measured by your side too.
Does SEO still pay now that AI answers the question?
The demand question got sharper, not obsolete. AI Overviews and chat assistants are very good at finishing informational queries — definitions, comparisons, how-tos — which is exactly the traffic a lot of content programmes were built on. If your plan was eighty percent explainer articles collecting top-of-funnel visits, the arithmetic has genuinely worsened.
What hasn't changed much: queries where the person needs a business, not a summary. "Orthodontist in Indiranagar", "GST filing software pricing", "buy running shoes size 11". An assistant answering those still has to name somebody, and being the named somebody requires the same work — a crawlable site, clear entity information, real citations from places the model trusts.
Anyone quoting you a precise percentage for how much traffic AI answers have taken is guessing; the public data is patchy and varies enormously by query type and industry. What you can do is check your own Search Console split: what share of your clicks come from queries an AI could finish in a paragraph? That number, not an industry average, tells you how exposed you are. See what AI Overviews did to organic traffic for the longer argument.
The practical answer for most businesses: shift the content mix toward the queries that need a supplier rather than an explanation, and stop measuring sessions. If you're still judging SEO by traffic in 2026, you'll conclude it stopped working right around the time it started delivering better-qualified visitors in smaller numbers.