Paste any page you own. We fetch it live, exactly as an AI crawler sees it, and score it in seconds. You get the number, the seven pillars, and a list of what to fix.
AIOInsights
An AIOTruth evaluation type, delivered through AIOInsights.
Does this page give an answer engine a reason to cite it?
It will never tell you your content “was written by AI”. No available detector can reliably prove authorship for edited, mixed, or ordinary business content.
The Content Credibility Check evaluates machine readability, authorship, sourcing, originality assets, specificity, style patterns, and update provenance from the page’s retrievable HTML. You get two separate readings: a Citation Readiness Score, and a Style Flag Risk level for the observable patterns automated detectors may react to. The evidence is shown for every finding.
Two questions, and only one of them pays
What everyone asks
“Is my content going to get flagged as AI?”
The wrong question. Every tool answering it with a percentage is selling a number it cannot defend, because no available detector can reliably prove authorship for edited, mixed, or ordinary business content, which is nearly everything on a real business site.
What actually pays
“Does this page give an answer engine a reason to cite me?”
Measurable from your own HTML, deterministically, with receipts. It is also what decides whether you show up when someone asks an assistant for a recommendation in your category.
Does Google penalize AI-generated content?
No. Google has said repeatedly that it rewards helpful content however it is produced. There is no blanket penalty for using AI tools, and an assisted page can rank first if it serves the reader better than the alternatives.
What Google does penalize is scaled content abuse: mass-producing unoriginal pages that add little value. That is a violation no matter how the pages were made, by a model or by a person.
One change is worth knowing. On 15 May 2026 Google widened the opening definition in its spam policies, which now read: spam is “techniques used to deceive users or manipulate our Search systems into featuring content prominently, such as attempting to manipulate Search systems into ranking content highly or attempting to manipulate generative AI responses in Google Search.” That last clause is new, and it brings AI Overviews and AI Mode inside the same policy (Search Engine Land). Google did not publish a new list of named tactics with it, and it did not say anything about how the change applies to pages published earlier. Anyone telling you otherwise is filling in the blanks.
So the risk was never that a machine helped you write. The risk is publishing thin, unoriginal, unsourced pages at volume, and that risk is measurable.
Why this is not an AI detector
Detection is tractable on pure, unedited model output. Ahrefs measured that 74.2% of new web pages contain some AI content while only 2.5% are purely AI. So the case where detection works is a rounding error, and the case where it coin-flips is nearly everything else.
On blended human and AI text, the condition Weber-Wulff and colleagues call the most common, measured detector accuracy sits near 50%. Sadasivan and colleagues proved the ceiling: as models improve, detector accuracy is bounded toward chance by construction. OpenAI withdrew its own classifier in July 2023 citing, in its own words, “its low rate of accuracy”.
The errors are not random either. Liang and colleagues found a 61% false-positive rate on essays by non-native English speakers, and simplifying native-speaker essays pushed their false-positive rate to about 57%. The bias tracks linguistic simplicity, not authorship. Plain writers sit in the machine region by construction.
If we sold you a percentage, the first client who asked us to prove it would be owed an apology. So we measure what we can show you instead.
What the check measures
Seven pillars, scored from your page's own HTML exactly as an AI crawler receives it, with no JavaScript executed.
- Machine readability. Is the text in the raw HTML at all? AI crawlers do not run your JavaScript, so if the words only appear after a script runs, an answer engine reads nothing.
- Authorship provenance. Is a real, named person credited in a way a machine can parse? A byline buried in prose does not count, because a crawler cannot tell it from a testimonial signature.
- Evidence and sourcing. Do claims link to primary sources, quote named people, and carry statistics that sit next to where they came from?
- Originality assets. Is anything here first-hand? Your own photographs, your own data, your own table. The part a text generator cannot fake for you.
- Specificity. Named places, real prices, actual dates, concrete numbers. The direct inverse of generic.
- Style sameness. Does it read as templated? This is where Flag Risk comes from.
- Update provenance. Are the dates honest, or has the page been stamped fresh without the words changing?
The weighting adjusts by page type. A homepage has no byline and no citations by design, so it is not scored as a failed article.
Why AI is not citing your site
When someone asks an assistant for a recommendation in your category and you are not in the answer, it is rarely mysterious. It is usually one of the seven above, and most often the same three: the page is anonymous, it sources nothing, and it says nothing only you could say.
Answer engines quote what is readable, attributable, sourced and specific. A page that is none of those gives an engine no reason to prefer it over the competitor who is.
Common questions
Why is my content flagged as AI when I wrote it myself?
Because detectors react to the shape of writing, not its origin. Uniform sentence length, narrow vocabulary and formal register all push text toward the machine region regardless of who typed it. Compliance copy, policy summaries and standard business writing get flagged for exactly this reason. It is not evidence about you.
How do I make my content sound less like AI wrote it?
Break the rhythm first, because uniform sentence length is the strongest surface pattern. Put a three-word sentence next to a thirty-word one. Then replace tell-words with the plain word you would have used out loud, let lists be the length the subject actually is instead of always three, and cut the hedging. Then add what no generator can invent for you: a named author, a primary source, an original number, a photograph you took.
Is the score deterministic?
Yes. The same page against the same rubric version always returns the same number. No model in the scoring path, no randomness. Every result is stamped with its rubric version, so a score can only move when your content changes or the published rubric changes.
Does this check my whole site?
No, it checks one page. That means it cannot measure phrasing recycled across your pages, and it says so in the result rather than passing you on a test it did not run.
What do I do with the result?
Fix the list, top down. It is ordered by how much each defect costs your score. If you would rather have it done as a system across every page, that is what Digilu does.
Want this fixed across every page?
The check is free. If the list is longer than you want to work through, Digilu builds and runs the whole system: the content, the sourcing, the schema, the trust signals, measured the same honest way.
Two levels of responsibility.
Credibility can change as content, authorship, evidence, sources, and public expectations change.
You can handle the recommendations yourself, have Digilu keep watching for changes, or have Digilu maintain the supported foundation. Nothing here expires and nothing changes if you do nothing.
We keep the foundation right.
Digilu brings your business into The Observatory, keeps watching the supported signals, and handles the supported maintenance of the website and core digital foundation.
- Continuous observation and meaningful-change alerts
- Digilu maintains the website and core digital foundation
- Entity details kept consistent across the sources you appear on
- Monthly Mission Briefing covering what changed and what Digilu completed
Secure Stripe checkout. Automated onboarding. Cancel any month.
Have Digilu manage your reputation
We manage your reputation and what happens after the sale.
Everything in Trust Foundation, and Digilu also manages your Google Business Profile, the reputation channels that decide whether people choose you, and the retention, referral and advocacy work after a customer buys.
- Everything in Trust Foundation
- Google Business Profile management
- Reputation-channel management and review stewardship
- Retention, referral and advocacy after the first sale
Secure Stripe checkout. Automated onboarding. Cancel any month.
Trust Foundation carries the observation and the supported maintenance of the website and core digital foundation. Adaptive Management carries all of that and additionally manages the Google Business Profile, the reputation channels that decide whether you are chosen, and the relationship after the first sale.
A result tells you where you stand. Digilu takes responsibility from there.
AIOInsights evaluates the public signals as they are today. The Observatory keeps watching them. Digilu carries the level of responsibility you choose, and no more than that.
Adaptive Management: We manage your reputation and what happens after the sale. · Trust Foundation: We keep the foundation right.
Check another part of the picture.
Content Credibility Check covers one dimension. These evaluate the others.