Citation-friendly pages
WEAKOnly /pricing matches a best-of, alternatives, comparison or review pattern. There's no "ImageVivid vs photoroom.com", no "best AI product photo tools" and no AI vs studio photography comparison.
3 pages analysed for AI-writing patterns.
54/100
MIXED
Scored 54/100 across 3 pages. The writing has some shape but several tells per page (16 flagged) suggest AI assistance without an editor's pass. Cleanup is straightforward, not a rewrite.
HOW THE SCORE WORKS
0-40
Human
Voice, specifics, edits visible.
41-60
Mixed
Some shape, several tells per page.
61-80
AI-heavy
Formulaic. Demoted on next core update.
81-100
Fully AI
Em dashes, stock openers, repetition.
01/blog/shopify/how-to-create-ai-product-photos-for-shopify1,383 words · 4 tells40
/blog/shopify/how-to-create-ai-product-photos-for-shopify
1,383 words · 4 tells
- OTHER
“The safe workflow is not “generate and publish.” It is: prepare an accurate reference, create controlled variations”
Negation-then-correction framing is a common LLM rhetorical scaffold, followed by a tidy rule-of-three list.
- REPETITION
“Compare it side by side with the real product and reject the image if any of these checks fail”
The guide repeats compare, reject and real SKU across nearly every section, which gives it a formulaic, checklist-heavy rhythm.
- OTHER
“Product color, shape, material, and labels are accurate”
Rule-of-three and four lists recur throughout (color, shape, material, labels), a generic LLM enumeration habit.
- STOCK PHRASE
“Treating white backgrounds as a universal Shopify rule”
Templated 'Common mistakes to avoid' headings with uniform one-line explanations read as generated structure.
02/blog/product-photography/gpt-image-2-5-flare-vs-sunburst-product-photos1,669 words · 6 tells55
/blog/product-photography/gpt-image-2-5-flare-vs-sunburst-product-photos
1,669 words · 6 tells
- NOT JUST X, BUT Y
“the useful distinction is not simply “cheap versus premium.” It is iteration speed versus edit precision”
Classic 'not X, it is Y' contrast scaffolding used to fake depth.
- OTHER
“More pixels do not repair an inaccurate product representation.”
Aphoristic, quotable one-liner closer typical of LLM prose.
- OTHER
“Reference preservation. OpenAI says subjects from reference photos remain more recognizable across new settings, styles, and compositions.”
Bolded-label bullet list with parallel structure and generic claims, formulaic LLM layout.
- OTHER
“Queue load, resolution, input count, prompt complexity, moderation, and upstream service conditions can all change the total wait.”
Over-long hedging list of factors, a rule-of-many pattern common in LLM text.
- STOCK PHRASE
“Reserve Sunburst for approved directions where preservation and focused refinement are worth more time and credits.”
Generic advisory phrasing and balanced recommendation closer typical of LLM output.
- OTHER
“These are product positions, not a guarantee that one model wins every prompt.”
Over-hedged disclaimer with a 'not X' contrast that recurs throughout the article.
03/blog/etsy/etsy-listing-photo-requirements1,575 words · 18 em dashes · 6 tells68
/blog/etsy/etsy-listing-photo-requirements
1,575 words · 18 em dashes · 6 tells
- EM DASH
“Etsy listing photos follow Etsy’s own rules — not Amazon’s and not Shopify’s.”
Unedited em dash in the opening sentence; the same clause separator recurs about 11 times across the page.
- EM DASH
“For the expanded 2026 workflow—including the current 20-photo and 2-video allowance”
Another em dash, this one unspaced, used as a parenthetical aside a human editor would turn into commas.
- EM DASH
“Shoot the first photo horizontal (landscape) or square — avoid pre-cropped squares that throw away context.”
Em dash used mid-bullet as a clause separator. A period or comma would do.
- OTHER
“It is not a generic photography tutorial and not a product pitch.”
Negation-pair scaffolding ("not X and not Y") that LLMs use to sound positioned. It also echoes the opening "not Amazon's and not Shopify's".
- OTHER
“Etsy’s visual culture rewards context and craft”
Abstract, vague claim with a stock-sounding rule-of-three cadence. It has no specific evidence behind it.
- OTHER
“treat the output as a draft, compare it against the real item, and reject anything that alters color, texture, lettering, or proportions”
Formulaic rule-of-three instruction followed by a list of four. The generic advice reads as machine-written.