Best search engine optimisation tools compared, ranked by what each one does that you cannot do by hand
A shortlist ordered on one checkable test: could a careful person produce this answer in an hour without the software? The tools that pass sit at the top, the ones that fail sit at the bottom, and one job on the list has no tool in this market at all.
Published by AI Knows Us · Updated 22 October 2026
There are about forty products sold as search engine optimisation tools and most buyers end up with three of them doing the same job. The way out is not a longer feature table. It is one question asked of each tool: what does this do that a careful person could not do by hand in an hour? Applied honestly it reorders the whole market, puts two free tools at the top, and leaves several well known products at the bottom with nothing to say for themselves.
The test is checkable because it has a yes or no answer for every tool, and you can run it yourself on anything not listed here.
The ranking, on that one criterion
| # | Tool | What it does that you cannot | Verdict |
|---|---|---|---|
| 1 | Google Search Console | Reports the queries Google actually recorded your pages appearing for. Nobody can reconstruct first party impression and position data from outside. Its help page, read 21 October 2026, documents a search type filter covering web as text based or multimodal, image, video and news. | Free, and first. There is no substitute and no paid product contains this data about your own site. |
| 2 | A third party link and keyword index: Ahrefs, Semrush, Moz or Majestic | Somebody else crawled the whole web and will tell you who links to your competitor. You cannot build this. Moz publishes Link Explorer at over 40 trillion links and a keyword index above 1.25 billion on its free tools page, read 21 October 2026. Majestic publishes its Lite account limits as 5,000 Site Explorer rows per search and 5,000 analysis depth, with no API access at that tier. | The one thing worth paying for, and one concentrated month usually answers the question. |
| 3 | A crawler: Screaming Frog, Sitebulb | Reads every page on your site and joins the findings. By hand this is a week. Screaming Frog publishes a free crawl limit of 500 URLs on its pricing page, read 21 October 2026, with scheduling, saved crawls, JavaScript rendering, crawl comparison and segmentation behind the licence. Sitebulb splits Desktop for local crawling from Cloud for scheduled crawling at scale, and its pricing page now advertises an MCP server that brings crawl data into Claude and ChatGPT. | Essential above a few hundred URLs, free below it. |
| 4 | Keyword Planner | Gives Google's own grouped demand estimate for a geography you select. You cannot guess this, and no free alternative is as close to the source. | Free with an Ads account. One figure per normalised group, so never add the phrasings inside a group together. |
| 5 | A rank tracker | Takes a clean reading from outside your own logged in browser, repeatedly, on a schedule. Doing it by hand is possible and doing it consistently is not. Semrush publishes 500 keywords tracked daily and 5 websites to monitor on its entry SEO tier, read 21 October 2026. | Worth it only when somebody reads the chart. Search Console answers the same question more honestly and later. |
| 6 | A content optimiser | Pulls the terms and subtopics common to the pages already ranking. You can do this by reading three pages, which takes about the hour. | Borderline. It saves time rather than answering something new. |
| 7 | An SEO score or grader | Nothing. It restates checks you can run in Search Console, PageSpeed Insights and the Rich Results Test, then adds a number nobody outside the tool recognises. | Last. A score is not a finding. |
Where this ordering disagrees with the market, and why
Three things fall a long way on this test. Keyword density tools fail it outright, because the calculation is arithmetic you can do in a spreadsheet and the number is not something search engines publish a target for. Site grader scores fail it because the underlying checks are free and public. And the AI writing feature attached to most suites in 2026 fails it in a different way: it does something you could do yourself, faster, which is a real saving, but it is not a measurement and should not be bought as one.
Two things rise. Search Console rises because people treat free as entry level and it is the only source of first party query data in existence. A third party link index rises because it is the one genuine information asymmetry in this market: the vendor spent years and a lot of money crawling the web and is renting you the result.
The buying order that follows from it
Verify Search Console and read the query report by landing page. Crawl the site free, up to 500 URLs, and fix what the crawl finds. Size the demand in Keyword Planner, remembering that one group gives one figure. Then, and only then, buy one month of one third party index, do the competitive work in a concentrated week, and decide whether a second month would tell you anything new. For a site under a few thousand pages it often would not.
The one check none of these tools runs, and who sells it
Every tool on this page measures a search engine. None of them measures the other place your buyer now asks, which is an assistant: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews compose an answer for each asker, naming some businesses and not others, and there is no index a crawler can read to find out whose name came up. That is the job with no tool in the list above, and it is not an oversight: the suites measure a crawlable index and an assistant answer is not in one.
AI Knows Us sells that one measurement and nothing else. We are not an SEO tool and we are not in the ranking above, because a buyer who came here for a crawler or a keyword index should buy one of the tools named above instead. What we do is ask your buyer question in all five apps on a schedule and record which businesses were named. Our own figures, from the rendered text of our llms.txt read on 21 October 2026: one credit is one buyer question on one app once, most checks cost one credit and a Claude check costs three because Claude costs more to run. The published US floor is $149 a month for SEO Start, and that same file says SEO Start watches 0 buyer questions on 0 engines, so it is a demand harvest rather than an engine watch. The first tier that watches the five engines is Growth at $299 with 100 buyer questions and 1,200 scheduled answers a month. A managed option is published in rupees only, Rs 88,499 to Rs 2,35,999 a month at a three month minimum, so we are not going to convert it for you.
Where the tools on this page beat us, plainly. Ahrefs Brand Radar, read on 21 October 2026, tracks prompts on ChatGPT, AI Overviews, AI Mode, Gemini, Perplexity, Copilot and Claude. AI Mode and Copilot are two surfaces we do not read at all, and its page publishes an overage price of $0.020 per check and the cost of a Claude update at 8 checks, which is a level of unit disclosure worth holding every vendor to, us included. Semrush, whose SEO and AI search pricing page we read the same day, sells AI visibility inside the suite you may already be paying for, with 50 prompts tracked daily and 300 AI visibility reports a day on its Starter tier. Moz now publishes AI Explorer, described on moz.com as showing what the top AI agents say about your brand. If you already pay one of those three, their module is the cheaper route and you should take it. We are the single purpose option for somebody who does not want a suite.
Before you shortlist anything, two things worth checking yourself
No competitor price table is reproduced on this page. What is quoted instead is the published limit that sets a tier, because a limit is the vendor's own statement of what you get and it does not move with the country you are reading from. That matters more than it sounds: on 21 October 2026 a fetch of one vendor's US pricing page from this machine returned Indian rupee prices, because the request left from an Indian address. A 200 response can still be the wrong page.
Also check that the product you are comparing still has the name you know it by. Ahrefs Webmaster Tools now says on its own page that it is Ahrefs Free. Amazon publishes that Amazon DSP has evolved into Amazon Ads Agent. A comparison built on a product name that moved last quarter is stale before you publish it.
If you want the same shortlist arranged by job rather than by this test: the SEO tool shortlist by job. On 21 October 2026 we read every link we could see on this site, 2,169 page bodies in all, against 2,082 live pages, and 935 of them, 44.9 percent, had no inbound link from any of those bodies. That figure is an upper bound rather than a count. No other page on this site links to the one above.
Questions people ask when shopping for an SEO tool
Do I need a paid SEO tool at all?
For technical work and your own search data, no. Search Console, a free crawl of up to 500 URLs and PageSpeed Insights cover it. You need to pay for exactly one thing, which is somebody else's crawl of the web so you can see your competitor's links.
Which one tool should a beginner buy first?
None, for the first month. Learn what Search Console says about your site, because every paid tool you later open will be interpreted against it. Then buy a month of one index.
Is an expensive tool more accurate?
For things it observes, no: a crawler either finds the broken link or it does not. For things it models, such as a competitor's traffic, more money usually buys a bigger sample rather than a correct answer. Treat modelled figures as comparative and never put one in a board pack as a measurement.
Why do two tools report different search volumes for the same phrase?
Because they are counting different things. Keyword Planner reports Google's own grouped estimate for the geography you selected. A third party reports a model built from its clickstream panel. Neither is wrong and they are not comparable.
How do I tell whether a tool is worth renewing?
Write down the decision you made last month because of it. If you cannot name one, cancel and keep the export.