SEO Audit Tool Buyer's Guide: 9 Questions to Ask

Published 2026-09-26 · 9 min read · by the isaeo team

The nine questions

Most audit tools demo identically: enter a URL, watch a progress bar, receive a long list of problems. The differences only show up in month three, when you find out whether you still open it. These nine questions surface those differences before you pay.

1. Does it rank findings by severity, or list them?
2. Can you explain the score to a client?
3. Does it separate real faults from things that don't apply?
4. Does it store previous crawls?
5. Can it alert you without you logging in?
6. What is the actual page limit per crawl?
7. How many sites, and what happens at the limit?
8. Are the reports something you'd send under your own brand?
9. Does it check AEO, or only classic SEO?

1–3: Does it prioritise?

Every crawler finds thousands of things. A tool earns its price by telling you which twelve matter this week.

Ask for the severity model. If findings aren't weighted, the tool has handed you the triage problem rather than solving it. A list of 1,400 issues is data; the top ten with severity and effort is an answer. See which issues actually move a score.

Ask what moves the number. A score you can't explain is unusable in front of a client, because the first question is always "why 64?" and "the algorithm" is not an answer.

Ask about false positives. This is the question that separates tools built by people who've done audits from tools built to a feature list. Concrete test: your login page is noindex, which is correct. Does the tool mark it down for having no meta description? It shouldn't — a page that has opted out of search shouldn't be charged for search-visibility issues. If every client site loses points for having a login page, the score is noise. We wrote up how scoring should handle this.

4–5: Does it remember?

Stored history. Without it you have a snapshot tool. With it you can answer "what changed since the last deploy?" — which is the question that actually prevents losses. Ask how many crawls are retained and whether you can compare any two.

Alerts you don't have to check. A tool that only reports when you visit it will catch a regression whenever you next remember to visit. Scheduled crawls plus an email on breakage is the difference between a report and a safety net — the whole argument in what not auditing costs.

6–7: Do the limits fit?

Pages per crawl. Check your real page count first — including pagination, filters and tag archives, which is usually several times what people guess. A tool that crawls 500 pages of your 3,000-page site produces a confidently incomplete audit, which is worse than none.

Sites, and what happens at the ceiling. Agencies hit this before the page limit. Ask whether adding site twenty-six means the next tier or a conversation with sales — and price the tier you'll be on in a year, not the one you'd start on.

Also ask what a "scan" counts as, because monthly scan quotas are where the real cost lives. Twenty-five sites crawled weekly is 100 scans a month, and plans quoting a generous page limit sometimes meter scans tightly.

8–9: Client-ready?

White-label output. If you serve clients, the report is your deliverable. Your logo, no vendor branding, a summary a non-technical reader can act on. A tool whose PDF advertises the tool is selling to your client over your shoulder. See white-label reports.

AEO coverage. Classic SEO checks are commodity — everything finds a missing title tag. Whether your pages are structured to be quoted by an answer engine is the newer and less-covered question, and it's increasingly what clients ask about. A tool that scores only classic SEO is answering half the brief; these are the AEO metrics worth having.

Red flags

• An issue count as the headline. "We found 2,847 issues!" is a volume boast. It usually means every missing alt attribute is one issue and nothing is weighted.
• A score with no published method. If nobody will tell you how it's computed, you can't defend it to a client or trend it honestly.
• "AI-powered" with no stated inputs. Ask what the model was given and what checks its output. Generated recommendations that reference pages the crawler never visited are worse than useless in a client deliverable — the validation matters more than the model.
• No free tier or trial with your own URL. You cannot evaluate any of the above from a demo of someone else's site.

That last one is the practical advice: run whatever you're considering against a site you know well. You already know what's wrong with it, so you can grade the tool's answer.

Buying well is half of it; using it well is the other half. See seven ways people waste money on SEO tools for the failure modes that come after the purchase.