Semrush Has Had Its Day

Updated 1 October 2026
I stopped rank tracking in two steps: the weekly check went in July, the monthly one on 8 September. Nothing I needed to know went with it. Google Search Console already gives me everything I need, and it comes from Google.

On 6 September the last scheduled rank sweep ran. It made 1,120 searches and cost about £12. Two days later I switched it off for good, and wrote the reason where the schedule used to be: keyword checks are ad hoc now, and Search Console is the main read.

Of the 255 phrases where the sweep found a page in the results, Search Console already had 251. It had them better, too.

Two different numbers

A rank tracker runs one search, from one location, at one moment, and reports what it saw as your position. Search Console reports where Google actually showed your page, averaged across every real search that surfaced it that month, with the impressions and the clicks that followed. One is a sample of a moment. The other is the record of what happened.

So chasing rankings every week means paying for a less accurate copy of a number you already own. I don’t see the point.

Semrush still sells its plans on how many keywords it tracks every day: 500 on the £113 plan, up to 5,000 on the £444 one (semrush.com, read 1 October 2026). I think tracking keywords daily is a thing of the past. Positions move gradually, so a daily number mostly records noise, and noise invites knee-jerk reactions that don’t help anybody. I’ve watched a tracker take one of my pages from 19 to 9 and back to 18 in a fortnight, and that 9 very nearly got published as a result.

A weekly ranking report reassures the person reading it that something is being watched. It doesn’t tell them what to do. Position 7 to position 5 on a phrase nobody searches is movement, not progress. The report that matters says what changed in the searches that bring customers, what it’s worth, and what I did about it. Search Console has all of that.

What Semrush actually gave me

So what could Semrush tell me that I don’t already have? Something I used to pay for deserves a fair hearing, and it comes down to three things.

Demand I don’t rank for yet. Search Console only knows about searches a site already appears in. If I’m deciding whether to build something new, I need search volumes and keyword ideas for a market I’m not in. That’s real, and Search Console can’t do it.

Other people’s sites. What a competitor ranks for, which searches they own that I don’t, who links to them. Search Console sees my sites and nobody else’s.

Reports that looked impressive. Domain scores, visibility charts, 40-row ranking tables with green and red arrows. This is where Semrush earned its place in agency life. It made a monthly report look like a month of work.

The first two cost pennies now

The first two gaps are questions, not subscriptions. I ask them when I’m about to build something, and the answer comes from DataForSEO, a pay-as-you-go data service billed per question. A full demand read, dozens of starting searches and a handful of competitors, cost me about 80p the last time I ran one.

That read is a tool I built into Westworld, the system that runs my business. It pulls the demand data, checks the competition, and comes back with a call: build it or don’t. The data backs up the call. Google Trends, which is free, is my cross-check on whether interest is rising or falling. Between them I get a better answer to “should I build this?” than a Semrush dashboard ever gave me, because the dashboard showed me numbers and left the decision to me.

Take the retired rank sweep out of this month’s bill and everything else came to about £4.

I used to pay a monthly subscription for a dashboard. Now I pay a few pence for an answer.

Semrush took 54 payments from me between October 2020 and May 2025, £4,538.95 in all. DataForSEO has had £244 from me since August 2025, paid as I go, and that has paid for the rank checks and demand reads I’ve run for my own sites and my clients’ since then.

What I paid each year: Semrush against DataForSEO 0 1000 2000 2020202120222023202420252026 £ a year 562 1486 1623 611 0 257 56 188 Semrush DataForSEO, paid as I go
What I paid each year: Semrush against DataForSEO Net payments from my company bank account, to the nearest pound. 2020 starts on 20 October; 2026 runs to the last statement on file, 22 September. The subscription ran to over £1,600 a year at its peak. Everything I've bought from DataForSEO since comes to £244. Company bank ledger, read 1 October 2026. Semrush: 54 payments, 20 October 2020 to 29 May 2025, £4,538.95 net of refunds. DataForSEO: £244, 20 August 2025 to 13 July 2026. No Semrush payments went through the account in 2024.
I didn’t import nearly 7 years of bank statements to win an argument about Semrush. I imported them to do my own books. The chart came free.

Paying for a question when I have one is a far better way to spend an SEO budget than paying every month for a dashboard in case I do.

Every new feature costs a new tier

Each time Semrush added something, it landed on a higher plan or as an add-on to the one you had. On 1 October 2026 the pricing page put keyword cannibalisation analysis, two of your own pages competing for the same search, on the Pro+ plan at £242 a month. Access to the raw data through the API is on Advanced, at £444. Even the reports are sold separately now.

On the same day I ran that cannibalisation check on my own site with Search Console and Claude Code. Three of my pages were competing with my homepage for the search I most want to win, “website design telford”, and the fix was live that morning. It didn’t cost a new tier. It cost a morning.

The third one I build better myself

That leaves the reports, and clients do want a report. They just don’t need Semrush’s.

Semrush will also, for an additional monthly fee, remove their branding from the reports you send to clients. You can pay them to hide the fact that you’re paying them. I found this instructive.

I don’t pay to hide a vendor’s name on my reports. There isn’t one.

My retainer clients get a report every month that Claude Code builds from their own Search Console data, written about their business and in their terms. It says what moved in the searches that bring them customers, what that’s worth, what I did about it and what happens next. It’s built for one business, so nothing on it is there to fill a page, and there’s no logo to pay to take off, because there’s no vendor in it.

Every one of those reports is a plain text file, kept alongside every report before it. Claude Code can read a client’s last 12 and tell me what’s moved, what keeps coming up, and what I said I’d do and haven’t done yet. You can’t compare 2 PDFs like that.

When a client does want a PDF, I right-click the file in Visual Studio Code and a free extension called Markdown PDF exports it. Semrush’s Base Report add-on, £8.10 a month, lists PDF export as a feature (semrush.com, read 1 October 2026).

Semrush was the right tool for its time: when the data was expensive to get, and when a report’s job was to look substantial. Neither is true any more. The data is pennies a question, the analysis is mine to do, and the report is built for the client instead of rented from a vendor. That’s why I think Semrush has had its day.


Read more about the practice:

If you want this kind of thinking applied to your own business — here’s how I work with clients, or get in touch.

← Back to Writing