AgentsBuilders Scanner tells you whether search engines and AI assistants can actually read, understand, and cite your page. Paste any public URL and it runs 135 checks across three categories — SEO, AEO, and GEO — then returns three scores plus an exact, ranked list of what to fix. There is no signup, no email, and no card required; however, a full scan still takes under 10 seconds, since every check runs against one parsed copy of the page.
Leave a contact and we will look at your site, then tell you what can be fixed automatically and what needs a human.
This scanner grades three separate things in one pass: SEO (can search engines rank the page), AEO (can AI assistants like ChatGPT and Claude parse and quote it), and GEO (would a generative engine choose to cite it in an AI Overview). However, the three scores often move independently — a page can pass nearly every classic SEO check while still failing most GEO checks, because GEO cares about things SEO never measured. In addition, all 135 checks run against a single parsed copy of the page, so no two checks can read the same fact and disagree.
SEO checks vs. GEO checks measure different things: one cares whether a crawler can index the page, the other cares whether a generative model would want to quote it. For instance, GEO looks for self-contained 60-180 word passages a generative engine could quote whole, original data phrased as "our data shows," and comparisons that read as balanced rather than one-sided. Our data shows the fastest points to recover are usually the cheapest to fix — a missing security header, a missing og:image, absent JSON-LD. Meanwhile, the slowest points to earn are genuinely original content, not padding written to satisfy a checklist. Finally, none of that matters if the scores contradict each other, which is why single-source parsing comes first.
Here's how to get scored in three steps — no account, results in seconds.
Any public page. No email, no card, no install.
135 checks run against the live page and return three scores.
You can start with whatever is ranked critical or high — every failed check comes with the exact fix.
Search has split into three channels. Therefore, we check all of them.
First, the fundamentals Google has ranked on for two decades: meta tags, headings, canonical URLs, structured data, broken links, security headers, and page performance.
First, can ChatGPT, Claude, and Perplexity even find your content. Then, can they parse and quote it — question-style headings, FAQ markup, direct answers, and authorship signals all matter here.
Finally, would a generative engine choose to cite you at all. That depends on original data, self-contained passages, trust signals, and topical depth.
135 checks total. Also, this is a real sample from each category — not a marketing summary, the actual check names from the scan engine.
Since each category (SEO, AEO, GEO) gets its own 0-100 score, it also gets its own letter grade.
| Range | Grade | What it means |
|---|---|---|
| 90-100 | A+ | Well optimized — keep monitoring |
| 80-89 | A | Also solid, with a few minor gaps left |
| 70-79 | B | Also workable, with room to grow |
| 60-69 | C | Also several real issues to fix |
| <60 | D / F | First, fix the critical items |
First, before shipping this page, we ran it through our own 135 checks.
The first scan of this exact page — before we fixed anything — scored 52% on SEO (31 of 54 checks), 60% on AEO (18 of 31), and just 37% on GEO (9 of 29), for an overall 50%. Our data shows the cheapest fixes were also some of the highest-value ones. Here's what that first scan actually caught:
We fixed what was honestly fixable and left the rest failing on purpose. For example, we won't invent social profiles we don't have, credentials nobody holds, or testimonials nobody wrote just to flip a red check to green — that would defeat the point of running an honest scanner. Instead, the score above reflects this page's real, current state; scan it yourself at any time to see today's live number, not a snapshot curated for this page.
Every check that reads a given fact about your page — the title, the robots.txt rules, the ARIA landmark count — reads it from one parsed source, shared by every check that needs it. Therefore, two checks can disagree about what matters, but they can never disagree about what is actually there in the HTML. That single-source design is what makes contradictory verdicts structurally impossible, not just unlikely.
We built this rule after finding a concrete counter-example elsewhere. Scanning a real site with another SEO/AEO/GEO scanner, its report contained two checks reading the same robots.txt and reaching opposite conclusions:
So we fetched the live robots.txt ourselves to check which verdict was right: it explicitly disallows GPTBot, ClaudeBot, Google-Extended, and CCBot. The FAIL was correct; the PASS on that same file, in that same report, was not. Meanwhile, the same report also disagreed with itself on ARIA landmark counts — 2 in one check, 3 in another — for the identical page, exactly the kind of self-contradiction a single shared parser is meant to prevent.
That is the failure mode this scanner is built to make
structurally impossible. A function called
findContradictions() runs on every single scan and
would refuse to ship a report where two checks read the same
source yet disagree — instead, it logs the conflict as an engine
bug for us to fix, not as normal output for a visitor to see.
If your page scores well on SEO but you have never checked AEO or GEO, you are optimizing for a search landscape that stopped being the whole picture years ago. Instead, run a scan, read the fixes ranked by severity, and start with whatever is marked critical or high — those move the score the most per hour of work. Also, if you would rather not do the fixing yourself, leave a contact in the form above once you have a report. Then, we will tell you honestly what is worth fixing and what is not.
No. The page is fetched once to run the checks, then discarded. Nothing is saved server-side. Instead, your scan history lives only in this browser's localStorage — clearing it removes it for good.
No signup, no email, no card. Paste a URL and scan.
Depth over breadth for this first version. Also, one URL gets all 135 checks properly instead of a shallow pass over many pages.
Not yet. Today this is diagnosis only. However, every issue includes a specific fix you or your developer can apply by hand.
Partly. The "what to fix" advice for each check is translated into all 7 languages. However, the diagnostic message itself (what's wrong, with the live numbers) still comes back in English — translating those too is planned as separate work.
About: this scanner is built and run by AgentsBuilders, a small team building AI agents that do real work. Read more about us.
Contact: questions or feedback — reach us through agentsbuilders.org, or use the "want this fixed" form above once you've run a scan.
Privacy: scanning a page fetches it once, server-side, to run the checks, then discards the HTML — nothing is stored. Instead, the only thing we persist is an anonymous, incrementing count of total scans (no URLs, no IPs) for the counter on this page. Scan history is kept in your own browser's localStorage, never sent to us.
Terms: this tool is provided free, as-is, with no uptime or accuracy guarantee. Also, scans are rate-limited per visitor to keep the service usable for everyone. Don't use it to scan pages you don't have permission to access.
Checks last updated .