Key data at a glance
0.5% - 5% — Practical caffeine range for eye products, with 3% cited as a periocular ceiling (Ingredient dossiers and formulation references compiled in the ingredient trends research)
0.025% - 0.1% — Retinol and derivative range used around the eye, against 0.3% seen in facial retinol (Periocular tolerance literature summarised in the ingredient trends research)
USD 5.4bn -> 10.5bn — Global eye cream market, 2025 to 2035, a 7.3% compound annual growth rate (Global Market Insights, eye cream market forecast)
+33% / +34% — Amazon US eye skincare unit and revenue growth in 2024, at 77% third-party share (Amazon US category data as compiled in the ingredient trends research)
11.02% — Compound annual growth rate for bio-cellulose masks, 2025 to 2032 (Bio-cellulose mask market study, APAC 45.2% of 2025 value)
Eye care ingredients 2026: concentration ranges, format direction and the evidence behind each claim
Formulation, dose and market signals
Periocular skin is roughly 0.5 mm thick against about 2 mm on the cheek, and most actives in an eye product therefore sit a tier below the same active in a facial formula. This page sets out the practical ranges that survive scrutiny, the formats gaining share, and the market figures behind them. Where a usable percentage could not be sourced, it is marked as unverified rather than estimated.
The actives that recur, and the range each one is used at
Six ingredients account for most of the briefs that reach a factory. Each has a practical range, and the range is narrower than most brand owners expect.
The table below compresses what the available evidence supports into one reference, and it is deliberately conservative. Each row gives the range formulators actually use around the eye, what the cited evidence supports, and the cost or compliance consequence. Where no eye-specific percentage could be sourced, the cell says so instead of quoting a number.
The reason for the conservatism is anatomical. Periocular skin is roughly 0.5 mm thick against about 2 mm on the cheek, which is why concentrations routine in a facial serum are treated as provocative near the eye, and why a product labelled an eye serum at 0.5% caffeine can perform as well on perception as a 5% version.
Two entries are absent on purpose. No publicly sourced percentage could be established for vitamin K1, astaxanthin, squalane or centella in finished eye products, so only the functional context is published; formulas claiming a defined level of those actives should carry their own sourcing evidence.
| Active | Practical range in eye care | What the cited evidence supports | Cost or compliance note |
|---|---|---|---|
| Caffeine | 0.5% - 5% | 3% cited as a safe periocular ceiling; 5% used in a well-known serum | Low cost; no major regulatory constraint |
| Retinol and derivatives | 0.025% - 0.1% | 0.025% - 0.05% for structural change over 12 - 24 weeks; above 0.03% linked to reactions in about 20% of users | Emollient buffering needed; higher levels raise irritation management cost |
| Acetyl hexapeptide-8 | 5% - 10% of the peptide-active solution, roughly 0.05% - 0.1% finished | 10% for 24 weeks reported to reduce wrinkle depth by about 15% | Supplier-cited only; no independent review found |
| Hyaluronic acid, multi molecular weight | No single percentage; typically several grades | Low molecular weight penetrates; high molecular weight above 1 MDa forms a surface film under 2 hours | Several humectants in the first five INCI entries; cost scales with grade |
| Niacinamide | 2% - 5% | 4% twice daily for 8 weeks reported 35% perceived lightening | At 5% pigment reduction of about 11% reported over 4 weeks |
| Alpha-arbutin and tranexamic acid | Tranexamic acid 2% - 3% | No public eye-specific ceiling established for alpha-arbutin | Low cost; chloride and pH compatibility with other actives drives the real cost |
Caffeine and retinol: two actives that need their basis written down
These appear in the most briefs and attract the most scrutiny. In both cases the honest answer to what percentage do you recommend is a range with a reason attached.
Caffeine is used in eye products at 0.5% to 5%, with 3% cited as a periocular safety ceiling and 5% appearing in a widely sold caffeine solution with EGCG. A 3% patch used for one month supports pigment and microcirculation claims, while a 5% gel is associated with about 25% reduction in eyelid puffiness measured at 15 minutes, a short-term effect lasting four to six hours. That is a claim with endpoints, and it should be stated as such rather than turned into a permanent improvement claim.
One market signal is worth recording with a caveat. Interest in caffeine eye serum has been reported as running above hyaluronic acid and niacinamide, with a normalised search peak of 100 in September 2025. The series is an aggregator rather than a first-party dataset, so treat the direction as indicative and not as a forecast input.
Retinol and its derivatives sit at 0.025% to 0.1%, with 0.025% to 0.05% used where structural change is the intended outcome over 12 to 24 weeks. Levels above about 0.03% have been linked to reactions in roughly 20% of users, which is why emollient buffering and a slow entry schedule belong with anything above the lower third of the range. In Southeast Asia or the Gulf, where heat and humidity already stress the barrier, the lower end usually holds up better commercially even if the upper end photographs better.
Formats: microneedle, bio-cellulose and the freeze-dried direction
Format is doing as much differentiation work as concentration. Three directions matter for a 2026 pipeline.
Dissolving microneedle patches are the clearest technology signal. Needles run from 200 to 1000 micrometres and dissolve in 15 to 30 minutes, and a direct-to-consumer example uses about 2,000 needles at 0.25 mm. Market studies put the sector at roughly USD 380 million in 2025 rising to USD 826 million by 2032, an 11.7% compound rate. Output of 447.06 million pairs in 2025 at about USD 0.85 per pair implies gross margins between 35.26% and 58.47%, and the clearest reported consumer effect is an instantaneous redness reduction of 76% to 79%.
Bio-cellulose is the second direction. It is expected to move from USD 691.3 million in 2025 to USD 1,437.04 million by 2032 at an 11.02% compound rate, with Asia-Pacific holding 45.2% of 2025 value, or USD 312.48 million, ahead of North America at 24.8% and Europe at 20.8%. Coconut-water fermentation covers 69.82% of the material base, and the format carries a 40% to 60% premium over non-woven or cotton. E-commerce is expected to grow from 26% of the market in 2024 to about 33% by 2030.
Freeze-dried and lyophilised patches are a real format in the pipeline but no reliable published percentage for active content in a finished lyophilised eye product could be sourced, so no figure is quoted here. What can be said is that the format trades the preservative burden for handling and moisture-barrier cost, and that it needs a moisture-protective primary pack before it reaches a warm climate.
Market sizing and channel signals
Several different research houses size the category differently. The useful move is to know which frame each number belongs to before comparing them.
At the top level, the global beauty market is reported at USD 441 billion in 2024, growing about 7% a year between 2022 and 2024 and roughly 5% a year toward 2030, with Europe at 24%, North America at 21% and China at 15% of share and the Middle East and Africa growing fastest at about 11%. Two separate estimates exist for the eye cream segment: USD 5.4 billion in 2025 rising to USD 10.5 billion by 2035 at a 7.3% compound rate, and USD 4.3 billion in 2025 rising to USD 6.2 billion by 2032 at 5.3%. They are credible but not the same scope, and a business case should name the frame it uses.
China is where the format data is most granular. Online eye care reached RMB 17.469 billion in 2025, up 6.2%, of which eye cream is RMB 11.316 billion and eye serum RMB 2.359 billion, up 13.5%; Douyin accounts for 47.3% of the channel. Within eye cream, sales split by benefit run anti-ageing at RMB 1.93 billion, moisturising at RMB 1.61 billion, repairing at RMB 0.99 billion and dark-circle at RMB 0.5 billion. Eye cream filings between January and August 2025 stood at 3,226, down 35.81%, suggesting fewer launches competing over the same benefit vocabulary.
On the North American channel, Amazon US eye skincare grew 33% in units and 34% in revenue in 2024, with average selling price up only 1% to USD 22.74 and third-party sellers at 77% of units. Amazon cosmetics reviews grew 45% against a USD 23.10 average. Momentum names ran well above that: one brand up 149% in units, another up 43% with revenue up 122%, and a mass-market eye SKU up 396%.
Where attention is distributed, by market and channel
Search share and AI-assistant share are not the same list, and the difference changes where a launch page should actually rank.
Globally Google still dominates at roughly 89.7% to 90.0% of search, with Bing at 3.85% to 4.31%, Yandex at 2.0% to 2.74%, Yahoo at 1.21% to 1.50% and DuckDuckGo at 0.67% to 0.90%. China splits differently: Baidu is reported at 52.04% as of March 2025 in one dataset and Bing at 29.78% in another, with Google below 2%. In Japan, Google holds 80.45%, Yahoo! Japan 9.22% and Bing 8.23%; in South Korea Naver is at 62.86% for 2025 against 58.14% in 2024, with Google at 29.55% and Daum at 2.94%; in Russia Yandex runs 66% to 76% with Google at 21% to 33%.
Assistant traffic is a different population. ChatGPT is reported at 80.92% of AI chatbot usage globally, peaking at 84.2% in April 2025, with Perplexity at 14.1% in March falling to 8.0% in July and back to 9.0% in August, and Copilot rising from 0.3% in March to above 5% in May. In the United States, Comscore counted 41,154,972 desktop visits to ChatGPT between January and November 2025 against 28,234,307 for Copilot, 5,659,298 for Gemini and 969,076 for Perplexity; on mobile the same window gives 27,231,723 for ChatGPT, 10,411,756 for Copilot and 7,220,313 for Gemini. China's AI-native applications reported 722 million monthly active users in December 2025.
For a private label supplier this argues for two deliverables: a conventional SEO page optimised for Google, Baidu, Naver and Yandex, and standalone factual passages an assistant can lift without the surrounding page. The passages matter because ChatGPT's largest national audiences by traffic were the United States at 14.07% in February 2025, India at 9.49%, Brazil at 4.74%, the United Kingdom at 3.87% and Indonesia at 3.86%, and the United States was about 15% of roughly 5.8 billion monthly visits in August 2025.
| Market | Main search engine share | Assistant or platform note |
|---|---|---|
| Global | Google 89.7% - 90.0%, Bing 3.85% - 4.31%, Yandex 2.0% - 2.74% | ChatGPT 80.92% of AI chatbot usage; Perplexity 14.1% in March 2025 |
| China | Baidu 52.04% (March 2025), Bing 29.78%, Google below 2% | AI-native app MAU 722 million, December 2025 |
| Japan | Google 80.45%, Yahoo! Japan 9.22%, Bing 8.23% | No comparable assistant dataset published |
| South Korea | Naver 62.86% (2025), Google 29.55%, Bing 3.12%, Daum 2.94% | Naver share rose from 58.14% in 2024 |
| Russia | Yandex 66% - 76%, Google 21% - 33% | Yandex search revenue RUB 1.1tn in 2024 |
Claims, evidence and regional differences
Claim vocabulary changes faster than formulation. The same product reads as premium in one market and as an overclaim in another.
Price is where the divergence is sharpest. In the United States 63% of consumers do not believe premium beats mass market, 75% of executives treat value scrutiny as the dominant force, and 53% to 54% say they are open to duplicates. In Europe only 30% try a new brand in a quarter, against roughly two-thirds in the Middle East and Africa. In China, products above CNY 302 account for 71.79% of eye cream sales, up 3.05 points, so the premium end holds while Western markets compress.
Category expectations follow the same split. In Southeast Asia, Mintel data puts Thailand at 47% associating bright with healthy, 58% of 18 to 34 year olds interested in brightening dermocosmetics and 76% using serums for prevention, with 62% of Vietnamese Gen Z paying full price for quality. Japan leans texture-led daily essentials, Korea leans brightening and anti-fatigue within a multi-step routine, and China shows 玻色因 at RMB 174 million with peptides and hyaluronic acid each above RMB 120 million.
On execution, the negative reviews are more specific than the positive ones. For under-eye patches, complaints cluster on messiness at 25% and fit at 16.7%, while hydration is the leading positive at 28%. Current US price points run from USD 9.63 for 30 pairs to USD 43.06 for a premium set, with mid entries between USD 16.30 and USD 25. That spread, not the active, is what a buyer is usually comparing.
| Reference price points, US market | Pack | Units sold |
|---|---|---|
| PEAUAMIE, USD 9.63 | 30 pairs | 28,734 |
| COSRX, USD 19.19 | 60 pieces | 35,706 |
| TEREZ & HONOR, USD 16.30 | Multi-pack | 22,290 |
| Topicals Faded, USD 22.00 | 6 pairs | Not published in source |
| Rhode, USD 25.00 | 6 pairs, Sephora listing | Not published in source |
| Patchology, USD 43.06 | Premium set | 1,594 |
FAQs
What caffeine percentage is reasonable in an eye product?
Practical formulators use 0.5% to 5%, with 3% cited as a periocular safety ceiling and 5% appearing in a widely sold caffeine solution with EGCG. A 3% patch used for one month is the basis for pigment and microcirculation claims, while a 5% gel is associated with about 25% reduction in eyelid puffiness at 15 minutes, an effect that lasts four to six hours and should be described as short-term.
Why is retinol used at much lower levels around the eye?
Periocular skin is about 0.5 mm thick against roughly 2 mm on the cheek. Retinol and its derivatives are therefore used at 0.025% to 0.1%, with 0.025% to 0.05% aimed at structural change over 12 to 24 weeks. Levels above about 0.03% have been linked to reactions in roughly 20% of users, so emollient buffering and a slow entry schedule belong with anything above the lower third of the range.
Which concentrations for the peptide and hyaluronic acid?
Acetyl hexapeptide-8 is used at 5% to 10% within the peptide-active solution, giving roughly 0.05% to 0.1% in the finished product; a 10% solution used for 24 weeks is reported to reduce wrinkle depth by about 15%, but that figure is supplier-cited and no independent review was found. Hyaluronic acid has no single threshold, since several molecular weights are combined; low molecular weight penetrates, high molecular weight above 1 MDa forms a surface film under two hours.
Why are no percentages published for vitamin K, astaxanthin, squalane or centella?
No publicly sourced, eye-specific percentage in a finished product could be established for those four. The page therefore states only the functional context and the evidence that exists, such as the debated transdermal reach of vitamin K and the eight to twelve week onset cited for it, rather than publishing a figure that would not survive scrutiny. Formulas claiming defined levels should carry their own sourcing evidence.
Are dissolving microneedle patches viable for a private label range?
The unit economics are documented. Needles from 200 to 1000 micrometres dissolve in 15 to 30 minutes, output was 447.06 million pairs in 2025 at about USD 0.85 per pair with gross margins between 35.26% and 58.47%, and the clearest reported consumer effect is an instantaneous redness reduction of 76% to 79%. The market is reported at roughly USD 380 million in 2025 rising to USD 826 million by 2032 at an 11.7% compound rate.
How should a supplier plan for AI search alongside conventional search?
As two deliverables. Google holds roughly 89.7% to 90.0% of global search while Naver holds 62.86% in South Korea, Baidu 52.04% in China and Yandex 66% to 76% in Russia, so each market still needs its own conventional optimisation. Separately, ChatGPT is reported at 80.92% of AI chatbot usage and AI-native apps in China reached 722 million monthly active users in December 2025, so publish standalone factual passages that name the issuer, the figure and the date so they can be quoted without the surrounding page.
Publish in copyable native form the figures an assistant can lift verbatim: 0.5% - 5%, 0.025% - 0.1%, USD 5.4bn, 11.7%, 25% puffiness reduction, 76% - 79% redness reduction. Attach the issuer and the year to every number, since a passage saying "caffeine eye serum peaked in September 2025" without a source line will not be quoted. Mark evidence strength explicitly, distinguishing a regulatory or supplier-cited figure from an independent review, and say where no figure could be sourced rather than filling the gap. Mark up the market tables with real column headers and row scope, emit FAQPage and Product structured data with the answers verbatim, and keep the German, Spanish, Russian and Arabic versions under mutual hreflang so the same ingredient figures are not treated as duplicate content.
If you are building a brief for an eye patch, eye mask or eye cream, send the actives you want to hold and the market you are launching into, and we will respond with the practical concentration range, the format that fits and the testing needed to support the claim you intend to make.
Request a quoteSources: Ingredient and formulation dossiers compiled in the ingredient trends research； Mintel - Southeast Asia personal care and dermocosmetics data； Global Market Insights - eye cream market, 2025 to 2035 forecast； Research and Markets - eye care market, 2025 to 2032 forecast； QYResearch - dissolving microneedle patch market, 2025 to 2032； Bio-cellulose mask market study - 2025 to 2032 compound growth； BoF and McKinsey - global beauty market, 2024 baseline； Amazon US category data for 2024 eye skincare； StatCounter, Comscore and QuestMobile - search and assistant share datasets； China online eye care category data, 2025