Seodar

Questions, answered plainly

Including the ones where the answer is no. A limit you find after paying is a refund; a limit you read here and buy anyway is a decision.

Accessibility and WCAG

The questions this gets asked most, and the ones where a careless answer is a legal problem for somebody.

No, and no tool can - including every tool that says it can. Automated testing reaches roughly a third of WCAG's success criteria: the ones with an unambiguous machine test, like contrast ratios, missing form labels and broken heading order. The rest need a person to judge them. In 2025 the US Federal Trade Commission fined an accessibility vendor $1,000,000 for advertising that its automated product made sites conform to WCAG. We find the failures a machine can prove, and we say which ones those are.
Because accessibility failures are overwhelmingly template failures. A contrast problem in a card component exists on every page built from that card, so auditing twenty product pages finds the same fault twenty times instead of finding twenty different ones. The sample takes one page per URL shape before it takes a second of anything, home page first - and every scan records how many pages it audited out of how many were eligible, so you always know what was and was not looked at.
The opposite, and the difference matters. An overlay is a script that tries to modify your site in the visitor's browser. This changes nothing on your site - it reports what is wrong so your developers can fix it in the source. Overlays are also the subject of the FTC action above and of sustained criticism from disabled users and accessibility lawyers, who point out that a script cannot decide whether alt text describes an image.
They may be right about the things they checked. Run the free scan and find out in about a minute - it costs nothing and needs no account. Our own site is a useful example: we shipped an accessibility page, then ran our own scan against our own site and found a WCAG 1.4.1 failure on 25 of 25 pages, including the home page. Links in body text that were distinguished from the sentence around them by colour alone. We fixed it the same day.
Same engine underneath - Lighthouse's accessibility section is axe-core, and so is ours. The difference is that Lighthouse tests one page when you press a button, and this tests a sample across your whole site on every scan, keeps the history, ranks the findings with your technical ones, and tells you when something new appears. If you only ever need one page checked once, use Lighthouse; that is a real answer and not a sales evasion.
axe-core 4.10.2, pinned by digest at build time, against the rule tags for WCAG 2.0, 2.1 and 2.2 at levels A and AA. Pinned rather than floating because a rule set that updates under you turns a clean scan into a regression nobody caused. Level AAA is deliberately not included - the W3C itself does not recommend AAA as a whole-site target.
The legal position depends on where you trade, and we are not the people to advise you on it. What is true regardless: the same requirements that make a page usable with a screen reader - real headings, labelled controls, text alternatives - are the ones that make it parseable by a search engine. The overlap is not a marketing angle, it is the same instruction written twice.

How it compares

Answered by what each tool is for, rather than by a feature table nobody believes.

Screaming Frog is a desktop crawler and it is excellent at being one - you run it, it crawls, you read the result. This runs in the cloud on a schedule, keeps every crawl so a change has something to be a change from, renders JavaScript where a page needs it, and joins the crawl to Search Console, Analytics, Cloudflare and Bing. If your work is a deep one-off audit on your own machine, Screaming Frog is probably the better tool. If it is watching a site over months, this is.
Semrush is a research suite: keywords, backlinks, competitor visibility, at a scale we do not attempt. What it does not hold is the response Googlebot was actually handed on your pages, whether Bing has your site in its index, or whether your pages pass WCAG. Those are different questions. Plenty of teams run both, and if you only want one, keep the one that answers the question you actually have this quarter.
Check counts are not comparable between tools and everybody in this market knows it - one vendor's single check is another's five. What is comparable is whether a tool shows you its arithmetic. Our score is 100 minus per-category penalties, each capped, each derived from a rate per page rather than a count, and every check is listed publicly with its severity and weight. If a number cannot be argued with, it cannot be trusted either.
Yes - CSV for findings and for pages, and a read-only API with an MCP server if you want it in code or in an assistant. Your data leaves as easily as it arrived, which is the only honest position for a tool you might one day stop paying for.

The health score

One number out of 100, and every question people ask about whether to believe it.

It starts at 100 and subtracts a penalty per category, each capped so no single kind of problem can sink the rest. A check's penalty is its impact multiplied by a severity weight, scaled by how many of your pages carry it - a rate, not a count, so a 5,000-page site and a 200-page one can be read against each other. The full arithmetic is on the report, not behind a support ticket.
Usually one of two things, and the report says which. Either the check catalogue grew - a new version measures more, so a score from one version is not subtracted from another, and the comparison refuses rather than pretending. Or the crawl covered a different number of pages: a 60-page crawl and a 500-page crawl of the same site reach different parts of it, so those are not subtracted either.
No. That was true until recently and it was wrong: on our own site, four fifths of the points we were losing came from configuration that was correct - noindex on sign-in pages, and the like. A check whose own description says the finding may be intentional now reports without scoring. Findings you disagree with should be visible and arguable, not silently priced in.
Yes, and they are deliberately kept out of the score. Your rules are yours; the number has to mean the same thing on every site or it means nothing. Custom findings appear in the report with everything else and carry zero weight in the arithmetic.

Data, access and privacy

What we hold, what we ask for, and what happens when you leave.

Read-only, and the narrowest scope each vendor offers. Cloudflare is a token you create yourself with one permission on one zone. Bing is a read-only API key from your own account. Search Console and Analytics go through Google's own consent screen. Every credential is encrypted before it is stored and none is ever shown back to you.
No. The AI features - rewriting a title, explaining a report in plain language - send the relevant findings to a model to produce that answer and nothing else. Your page content is not used to train anything, by us or by anyone.
In the EU. The crawler identifies itself as SeodarBot with a published user agent, respects robots.txt, and signs its requests so a site owner can verify that traffic claiming to be us actually is.
Disconnecting takes the data with it - the rows that integration filled are deleted, and revoking a Google source removes what it produced. A connection left unattached to a site is purged after thirty days, and you are told when that happens rather than discovering it.

Plans and billing

One site, crawls up to 200 pages, the full check catalogue, and the accessibility pass - which is not held back for paid plans. No card. The free scan on the home page goes further still: 25 pages with no account at all.
One credit covers 10 crawled pages, and credits are only debited when a crawl finishes. A crawl that fails costs nothing. The AI features spend credits too: 1 for a rewrite, 3 for a summary, and nothing at all when the answer has not changed since you last asked.
Yes, from Stripe's own portal, and the plan runs to the end of the period you paid for. Nothing is deleted the moment you downgrade - you drop to the free plan's limits rather than losing your history.
Neither. Both are on every plan including Free. Bing costs us nothing to offer - the API key is yours and the quota it spends is yours - and accessibility is the one thing in this category nobody else ships, which makes it worth more as a reason to try us than as a reason to pay more.

Questions about one feature

Each feature page answers its own, beside the thing it describes.

Something not answered here

Ask, and the answer goes on this page rather than into one inbox. Get in touch