Can a tool tell you whether your website looks AI-generated?
Yes, but only in a narrow sense. A tool can count measurable patterns that AI-generated pages often share: em-dashes in copy, filler marketing verbs, placeholder names and stock purple gradients. GROX describes such a check as counted facts rather than a model's opinion.
Why do counted patterns matter more than a model's opinion?
A model's opinion is hard to act on. If a report says a page feels generic, the owner is left guessing which sentence or colour choice caused it. A count is different: it points to an exact location. Each em-dash is an em-dash. You can search for them, delete them or keep them because your style guide asks for them.
Counted signals also make before-and-after comparisons possible. After an edit, the number either falls or it does not. There is no hidden reasoning to untangle. The GROX site check applies the same logic: it counts the measurable tells alongside SEO checks, never a model's opinion.
| Pattern | What gets counted | Why it helps | When to ignore it |
|---|---|---|---|
| Em-dashes in copy | Instances of the em-dash character in visible text | Every instance can be found and edited directly | A house style may use them deliberately |
| Filler marketing verbs | Verbs that add tone but no information | Shows which sentences may be shorter without losing meaning | Some industries use them as normal phrasing |
| Placeholder names | Names that read as stand-ins rather than real people or companies | Flags unedited template text | A legal or demo page may need generic names |
| Stock purple gradients | Colour transitions common in AI-generated design | Highlights a design choice you can replace or defend | The brand may already own that palette |
What tells are worth checking first?
Start with the one that is easiest to verify. Search the rendered page for the em-dash character. If the count is high, read each instance and decide whether it is doing a job that a comma or full stop would not do better.
Then look for filler marketing verbs. These are words that create motion or excitement without adding facts. Removing them often shortens a sentence and makes the offer clearer. The check can locate them; you still have to judge whether the sentence improves.
Placeholder names are worth a pass on any page that is supposed to show real customers, team members or case studies. If a name looks like a stand-in, either replace it with a real one or mark the page as illustrative.
Purple gradients are the least urgent. A gradient is a legitimate brand choice. If it matches your palette, keep it. If it was selected by default, changing it is a quick visual correction.
- Search for em-dash characters and review each one.
- Remove filler marketing verbs that add no facts.
- Replace placeholder names on pages that claim to show real people or companies.
- Review purple gradients against the brand palette before changing anything.
What do these terms actually mean?
The terms below are useful because they can be counted or observed without asking a model for a subjective opinion.
- Em-dash
- A long dash used in place of commas, colons or parentheses. It is a common pattern in AI-generated copy and easy to find with a text search.
- Filler marketing verb
- A verb that adds tone but no information. If the sentence still works without it, the verb is filler.
- Placeholder name
- A name that reads as a stand-in rather than a real person, product or company. It often appears in template or sample copy.
- Stock purple gradient
- A colour transition using purple tones that appears frequently in AI-generated interface designs. It is a pattern, not a defect.
When is a simpler check the better choice?
For a small site, a manual pass may be enough. A find-and-replace for em-dashes, a short list of filler verbs and a quick look at the brand palette cover most of the ground without any new tooling.
A count-based check earns its place when there are many pages, frequent edits, or several people publishing without a shared style guide. It gives every page the same baseline review and records whether a fix held.
The main trade-off is that counting finds patterns, not outcomes. A page can pass a pattern check and still convert poorly. A page can contain em-dashes and still read as the brand's voice. Use the check as an editor's assistant, not as a replacement for user testing or a careful read.
Common questions
Does an em-dash in my copy mean my website looks AI-generated?
No. A single em-dash is not evidence. A count-based check flags how often the character appears, and you decide whether each instance fits your brand. Some teams use em-dashes deliberately in a house style. The value is finding every occurrence quickly, not treating one mark as conclusive.
What are filler marketing verbs?
They are verbs that add tone but no information. A sentence can often be shortened by removing them, and the meaning stays the same. Because they are countable, a check can point to the sentences most likely to benefit from a rewrite, rather than asking you to re-read the whole page.
Can a tool spot placeholder names?
It can flag names that read as stand-ins rather than real people, products or companies. The check is pattern-based, so it may catch a name that looks like template text. Only you know whether the name is real or intentionally generic for a legal or demo page.
Should I remove a purple gradient just because it was flagged?
Not automatically. A stock purple gradient is a pattern common in AI-generated design, but it is not always wrong. If the gradient already matches your brand palette, keep it. If it was selected by default and conflicts with your brand, replace it. The flag is a prompt to review, not a rule.
If you want a count of these tells on your own site, run the GROX site check and see which patterns are actually present.