Copyright notices such as “© 2025 Company Name. All rights reserved” are widely used on commercial websites, despite their formal legal necessity being eliminated under international law. This paper investigates the actual value of such notices in legal, UX, behavioral, and marketing contexts. Based on an empirical study of 347 websites, we identify patterns in notice content, structure, and placement, as well as mismatches with legal ownership data. We analyze the evolving relevance of copyright notices under EU law, particularly in the context of Article 4 of Directive 2019/790 and text and data mining (TDM). The results reveal that copyright notices still serve multiple meaningful functions. Our findings indicate that while the traditional legal purpose of notices has diminished, they may still function as signals with implications for, among other things, compliance and innovation, highlighting the need for further research on their role in the context of generative artificial intelligence (AI)and large-scale TDM.
Contemporary websites act as digital showcases for organizations and individuals, as well as platforms for communication and information exchange. Content creators or rights holders frequently include unilateral copyright statements, typically in the lower section of a page (the footer), in a conventional format such as: [copyright symbol] [owner’s name] [year] [rights statement].
The phrase “All rights reserved” originates from the Buenos Aires Convention of 1910, which required this wording to formally reserve rights. Interestingly, the convention was signed only by the United States and several Latin American countries. After those countries acceded to the Berne Convention in the 1970s, the formal requirements of Buenos Aires lost their legal significance – under the Berne Convention, copyright protection arises automatically upon creation and requires
Despite that, the statement is still widely used today. What value does it provide to rights
holders? Some scholars argue that the © symbol functions as a deterrent, fulfilling a behavioral and
social role [
On the one hand, such “little texts” are often not designed for active engagement. They present
demands without explanation, restrict user/AI action without interaction, and embed legal
requirements into visual noise [
Research suggests that the effectiveness of a notice depends less on format (e.g., symbol,
structure, or phrasing) and more on clarity, contextual relevance, and user perception [
Thus, the scholarly view affirms the prevailing behavioral value of copyright notices. Copyright
is also an engine of standardization – by shaping routines and expectations, it facilitates mutually
beneficial transactions in digital markets [
To address that question, we now examine the frequency and quality of copyright notices in real-world website practice. This paper critically examines why copyright notices like “All rights reserved” remain widely used on websites and what functions they fulfill in the digital environment. In the current era of rapid AI development, TDM has become a central mechanism of innovation and knowledge extraction. The capacity of algorithms to learn from vast corpora of online texts and images has transformed the very meaning of copyright boundaries. Within this new context, the simple textual notice “All rights reserved” gains renewed significance: it may determine whether content can lawfully be used in AI training datasets. Thus, examining how such notices function and are perceived today is essential for understanding the evolving relationship between copyright law, digital design practices, and the governance of generative AI.
The methodological framework (Figure 1) outlines the sequential stages of data selection, coding, and analysis used to identify functional patterns in website copyright notices.
To interpret the different patterns and functions of copyright notices, the empirical part of the
study was based on the Forbes Global 2000 list (2024 edition), which served as the initial
population. We selected websites of the world’s largest companies because their online practices
are often considered benchmarks in design [
Data collection was conducted in June 2025. We adopted a static snapshot approach to capture dynamic website content at a fixed point in time, enabling cross-sectional analysis.
We prioritized global .com domains when multiple regional versions were available, reflecting the international scope of the corporations and their preference for globally consistent branding. This choice may also indicate a strategic intent to reduce perceived dependence on national legal systems. However, such prioritization may underrepresent country-specific legal nuances and regional adaptation of copyright notices. Therefore, while the study reflects only a limited regional context, it demonstrates an internationalized approach to copyright notice practices.
No mobile versions were included.
The final sample consisted of 347 websites. Each homepage was manually examined to document the presence, structure, and content of copyright notices. The analysis focused on five predefined elements: (i) the © symbol, (ii) the word “Copyright”, (iii) the phrase “All rights reserved”, (iv) the year or range of years, and (v) the name of the copyright holder. To ensure systematic data collection, we examined website footers and (for other cases) performed on-page keyword searches using the following terms: ©, “copy”, “rights”, and “reserved” without spaces.
We applied a deductive coding approach based on these elements, iteratively refined during data
collection. Coding and documentation were performed using spreadsheets [
We also recorded additional features where present, including references to other forms of intellectual property, licenses, explicit mention of affiliated entities within the copyright notice etc.
To assess ownership transparency, we matched the named copyright holder on each website with the actual trademark owner. Trademark ownership data was retrieved from public database “TMview”. In our trademark search strategy, we applied a three-step filtering sequence: (1) search for trademarks matching the company name from the Forbes Global 2000 list; (2) filter by the company name; (3) filter by country of registration, corresponding to the headquarters' country (if needed). The process was iterative and continued until a legal entity name could be confidently established. Records associated with individuals or lacking clear links to the company were excluded.
While trademark ownership does not perfectly reflect copyright ownership, for large multinational corporations these portfolios are often centrally managed by the same legal entities, making trademarks a reasonable proxy for identifying the corporate structure of rights holders. This approach, however, introduces certain limitations, as discrepancies may still occur in cases where copyright is held by subsidiaries or external contractors.
An exploratory subgroup analysis was conducted for companies headquartered in the European Union (n = 54). For these, a longitudinal review using the Internet Archive’s Wayback Machine was conducted to compare the structure and wording of copyright notices across three regulatory periods: (1) before the adoption of Directive 2001/29/EC, (2) between Directive 2001/29/EC and Directive 2019/790, and (3) after Directive 2019/790, based on the 2025 snapshot.
Content analysis focused on structural patterns. Our method was recursive and reflexive, allowing adjustments during the research process. This included repeated cycles of sampling, analysis, and comparison, particularly useful given the dynamic nature of websites.
The methodological approach draws upon classic principles of snapshot analysis in dynamic
web environments [
The study employed descriptive statistics and qualitative interpretation to identify prevailing patterns and deviations in copyright notice design. Based on the total population of 2,000 companies. We computed the margin of error using the finite population correction: e=Z∗√(( p∗(1− p)/ n)∗(( N −n)/( N −1))) (1) where =1.96 – critical value corresponding to a 95% confidence level, p = 0.5 – assumed maximum variability (most conservative estimate), n = 347 – sample size used in the study, = 2000 – total population size (Forbes Global 2000 list).
This yields = 0.048 (±4.8%), confirming that the sample achieves the standard 95% confidence level.
Several limitations must be acknowledged. First, the study analyzed only the main publicly accessible version of each corporate homepage, primarily in the .com domain. As such, regionspecific legal disclosures or adaptations on localized versions (e.g., .de, .jp) were not captured, potentially omitting local legal compliance nuances. Second, while data collection occurred at a specific point in time (June 2025), websites are dynamic media. Although screenshots were not taken during data collection, the findings remain verifiable through archival tools such as the Internet Archive’s Wayback Machine. Finally, although efforts were made to verify the legal entity listed in notices through trademark databases, the accuracy of associations may be limited by database completeness and naming discrepancies.
Among the analyzed companies, 8% of homepages did not contain any copyright notice; statistics for those that did are presented in Table 1.
To quantify the uncertainty of this estimate, we calculated a 95% Wilson confidence interval for the observed proportion of websites using the phrase “All rights reserved”.
pW =( p +( z2/(2∗n)))/(1+( z2/ n)) (2) SEW =( z /(1+( z2/ n)))∗√(( p∗(1− p)/ n)+( z2/( 4∗n2)))
CI = pW ± SEW (3) (4) where p = 0.47 – observed proportion of websites containing the phrase “All rights reserved”, p = 0.5 – assumed maximum variability (most conservative estimate), z = 1.96 – critical value corresponding to a 95% confidence level, CI is [0.42, 0.52] – resulting lower and upper bounds of the confidence interval.
This indicates that between 42% and 52% of corporate websites in the population are expected to include the phrase “All rights reserved.” Formulas (2)–(4) provide statistical validation by establishing a Wilson confidence interval for the observed frequency of “All rights reserved,” ensuring that the measured prevalence of this phrase across websites reflects a reliable and interpretable range rather than a point estimate.
Among the copyright notices that did include a date, the formatting varied. Table 2 summarizes the types of date formats used and their respective frequencies.
The identification of rights holders in copyright notices revealed significant inconsistencies. While most websites included a name, discrepancies between the listed entity and the actual legal owner were common. In 283 cases, we were able to use a trademark database to identify the legal owner. Table 3 summarizes the accuracy and types of name attribution found in the notices. Trademark / website / other name instead of legal entity
Four companies listed multiple alternative legal entity names. Some (5%) companies (mostly Chinese) included (ICP) license numbers. In seven cases, affiliates / subsidiaries were specified.
In nine cases, the notice additionally referred to trademark rights, in three cases – to patents; in one case – to industrial design rights.
A k-means cluster analysis revealed four dominant structural configurations of website copyright notices (Figure 2). The minimalistic cluster, characterized by the use of only the © symbol, reflects the UX-related convention of visual simplicity and professional standardization. The traditional cluster, featuring the full formula including “All rights reserved,” aligns with the legal function of explicit rights reservation. The brand-focused cluster, which combines the © symbol with a trademark or brand name, embodies the marketing and reputational function of copyright notices. Finally, the hybrid cluster merges legal and branding elements, corresponding to the mixed functional type. These results indicate that the structure of copyright notices reflects deliberate organizational choices balancing legal signaling, user experience, and brand communication goals.
Although copyright notices are primarily designed for human users, the analysis shows that most are technically machine-readable due to their placement in HTML footers and use of standard text encoding. This means that, from a technical standpoint, such notices can already be detected and parsed by automated systems without requiring additional metadata formats. Consequently, plain-text copyright statements may function as a basic layer of machine-readable rights reservation, bridging legal intent and digital accessibility. This finding reinforces the technical feasibility of textual copyright reservations as practical indicators for automated TDM compliance and AI dataset governance.
To examine the historical evolution of copyright notice practices, a longitudinal review of corporate websites was conducted for those associated with companies headquartered in the European Union. Out of the full sample, 54 websites were linked to companies headquartered in the EU. Using the Internet Archive’s Wayback Machine, we analyzed archived versions of these websites at three key stages:
Before the adoption of Directive 2001/29/EC, Between Directive 2001/29/EC and Directive 2019/790, and
After the adoption of Directive 2019/790, based on the current dataset.
Archived versions were successfully located and analyzed for 48 websites in the intermediate period and for 32 websites in the pre-Directives period.
Table 4 provides a comparative overview of copyright notice features across these three stages.
As the results show, the inclusion of a copyright notice is a prevailing norm (present on 93% of websites examined). However, the content and format of such notices vary significantly.
This diversity reflects the fact that copyright notices may serve multiple, partly overlapping functions, which can be grouped into four broad categories: legal; behavioral; UX-related; marketing or reputational.
As the four hypotheses represent conceptual typologies rather than statistically independent variables, their validation in this study is interpretive rather than inferential.
From a legal perspective, the notice may help establish willfulness in cases of infringement (17
U.S.C. § 401(d)), or act as a rights reservation signal under Directive 2019/790 in the EU. From a
behavioral standpoint, the notice can serve as a deterrent [
These hypotheses are not mutually exclusive: a single notice may simultaneously fulfill all four functions. In fact, the strength of the copyright notice lies in this multi-functionality, allowing it to operate across different domains (law, behavioral economics, and branding etc.). However, there may be tensions between these roles, for example, detailed legal language may conflict with UX clarity, or visual compactness may limit legal precision.
The frequency of use confirms the UX-related value and the established convention of placing the
notice on a website. A copyright notice is perceived as a marker of a well-maintained website,
reliable, and, conversely, a well-designed website is often perceived as an attribute of a credible
rights holder – this connection has marketing implications, as visual design is known to influence
user trust [
From a technical perspective, modern machine learning pipelines depend on large-scale automated extraction of online content, making the consistency and clarity of copyright notices crucial for lawful data governance.
Outdated years may negatively affect user perception and reputational value: users often infer
ongoing site maintenance from a current year indication, affecting perceived credibility [
It has also been suggested that, while designers often feel disconnected from legal language [
Interestingly, more than quarter of websites redundantly included both the © symbol and the
word “Copyright”, even though either form is widely understood. This redundancy may reduce
communicative efficiency [
The well-known rules state that increased textual density correlates with information overload
[
As for the claim that the behavioral function of copyright notices is paramount, we raise several objections. First, this assumption implies that all other functions are secondary or insignificant – we challenge it below. Second, effective behavioral regulation requires directive force, even if they do not follow the syntactic or pragmatic form of prototypical directives [23] – conditions that are rarely present in standard notices. Third, placement in the footer alone does not enhance its normative power. While we acknowledge the behavioral function, we do not agree that it dominates all others.
We now turn to the legal interpretation of the notice as a whole and its elements.
First, the persistent inclusion of copyright notices may reflect a misunderstanding among
designers and content creators. As noted by Yezril [
Second, some creators and rights holders appear to treat the notice as a catch-all form of protection, extending not only to copyright but also to trademarks, patents, or design rights. This is conceptually flawed: each form of intellectual property has distinct legal mechanisms and does not require a unified declaration. However, we observed that half of the notices included a commercial name (often a trademark) instead of a legal entity, and some explicitly referenced trademark, patent, or design rights. This indicates a blurring of legal categories, which may be motivated more by branding strategy [24] than legal accuracy. And the copyright notice may be a natural place to declare this.
Although some commentators previously argued – for example, Goryunov [25] – that copyright notices on websites have no legal significance, such a claim can no longer be accepted in full.
Article 5 of Directive 2001/29/EC on the harmonisation of certain aspects of copyright and related rights in the information society contained certain permissions regarding the reproduction in the press, communication to the public, or making available of published articles on current economic, political, or religious topics or broadcast works or other subject matter of the same character, in cases where such use is not “expressly reserved”.
The preamble to that Directive states that: “Technological development will facilitate the distribution of works, notably on networks, and this will entail the need for rightholders to identify better the work or other subject-matter, the author or any other rightholder, and to provide information about the terms and conditions of use of the work or other subject-matter in order to render easier the management of rights attached to them. Rightholders should be encouraged to use markings indicating, in addition to the information referred to above, inter alia their authorisation when putting works or other subject-matter on networks”.
The current Directive 2019/790 on copyright and related rights in the Digital Single Market no longer contains such a preamble, but it introduces an important Article 4, which could revive the legal relevance of using notices on websites: Member States shall provide for an exception or limitation in respect of the reproduction and extraction of works and other subject matter to which they have lawful access, for the purposes of TDM; that exception or limitation shall apply on condition that the use of works and other subject matter referred to in that paragraph has not been expressly reserved by their rightholders in an appropriate manner, such as machine-readable means in the case of content made publicly available online.
It reinforced this logic in the context of TDM, introducing the concept of “appropriate reservation”, and importantly, referencing machine-readable means.
The mentioned article from Directive 2019/790 has been transposed into the national laws of EU Member States [26], which became an adequate response to current technological developments.
However, for example, in the United States, there are no such provisions in the law, and therefore the legal weight of the notice is more limited, but it still exists – the fact of public access and the indication of the owner of the work and/or the date of its publication. As we noted before, in U.S. law, a copyright notice may also serve an evidentiary function, particularly in establishing willfulness.
In the context of aligning Directive 2019/790 with the subject of this study, the following question remains controversial: can the phrase “All rights reserved,” placed for example in the footer of a webpage, be considered an appropriate form of such reservation?
The phrase “All rights reserved” is presented as plain text on an HTML page, which makes it accessible to any automated analysis system. Modern TDM tools are capable of extracting text from HTML, PDF, OCR-processed images, and other sources. From this perspective, the “machinereadable” requirement is formally fulfilled.
In the context of AI, copyright notices can act as a primary signal for automated systems that scrape, index, or train on web data, determining whether such activities fall within lawful TDM exceptions.
The phrase “All rights reserved” has a well-established legal meaning in copyright law. It clearly signals that the author does not grant permissions beyond those explicitly stated, and therefore includes a prohibition on TDM unless otherwise specified.
If the phrase “All rights reserved” is placed in the footer of a website, it can be regarded as part of the terms of use of the site according to general legal understanding. The footer is a publicly accessible component of a website where legal disclaimers are traditionally located. Accordingly, a phrase in the footer can be characterized as an element of the conditions of access to the content.
Unlike links to licenses, external policies, or RDF (Resource Description Framework)
descriptions, “All rights reserved” is understandable to both humans and machines without the
need for additional uniform resource identifiers or formats. Legally meaningful conditions can be
formulated simply as a phrase in the footer [
The placement of a notice in the footer remains a subject of debate, as it does not always convey a clear or intentional legal meaning to human users [27], but it be interpreted differently by a machine.
The technical meaning of “machine-readable” in the context of the Directive 2019/790 remains contested. Modern forms of TDM opt-outs face the lack of a universally accepted machine-readable standard [28–30].
“Machine-readable” may include meta tags (<meta>), robots.txt files, RDF-formatted licenses, or rel="terms" attributes in HTML.
A phrase like “All rights reserved”, without specifying context (e.g., prohibition of TDM in particular) and without linkage to conditions (terms), may not allow a machine to definitively determine which actions are prohibited. This is especially important when a rightsholder may wish to prohibit some actions (e.g., commercial use) but not others (e.g., scientific TDM).
The EU aims to create an environment where automated systems can interpret the legal status of content without human legal intervention. Text in the footer (even if technically readable) is not standardized, may be stylistically formatted, is not linked to specific objects (e.g., only videos, only text), and is not always technically accessible to parsers if embedded in dynamic or non-semantic page elements. But the effectiveness of a notice depends not only on its placement, but also on the ability of platforms or AI systems to recognize and interpret its legal significance [29].
As generative AI models increasingly rely on massive web-scale datasets, the presence or absence of clear copyright notices directly shapes the legality of data acquisition for training, especially under emerging TDM-focused regulations. The Directive 2019/790 asserts that machinereadable methods may be considered an “appropriate reservation” in relation to publicly available online content. This may, in some jurisdictions, limit the legal weight of traditional textual disclaimers if not formatted in a way suitable for machine processing. A significant regulatory gap exists between the nominal rights established under EU legislation and their actual technological implementation [30].
Even greater confusion arises from European court decisions. Across EU Member States, there is divergent judicial practice regarding what constitutes an appropriate reservation of rights in order to exclude the exception provided by Article 4 of Directive 2019/790.
In Germany (Kneschke v. LAION), the court recognized that a textual notice on a website may be sufficient, provided that it clearly expresses a prohibition. This increases the value of a copyright reservation notice.
In the Netherlands (DPG v. HowardsHome), the decisive factor was the directionality of the notice, which diminishes its legal weight unless accompanied by additional statements.
The Hungarian Court of Appeal in Case 9.Pf.20.353/2024/6-II, which concerned TDM and the relevant provision of the Directive, did not contribute to the legal interpretation of the question at issue.
The position of courts in EU Member States indicates that the legal effectiveness of rights reservation under TDM depends not only on technical structure, but also on clarity and directionality.
The phrase “All rights reserved”, despite its clear legal meaning and technical accessibility to text-reading systems, may not formally meet the requirement of “machine-readable means” in the narrow sense used in EU legislation. Nevertheless, the position that such a phrase (especially when placed in the footer or within the terms of a website) expresses the intent of the rights holder remains logical and doctrinally sound.
A clear textual notice may increase the legal plausibility of enforcing rights under Article 4 of Directive 2019/790, its effectiveness remains conditional. Without structured formatting and standardization (as our empirical data show), its probative value is limited, but not null. Therefore, such notices may increase the rightsholder’s ability to hold unauthorized users liable, though this effect is mediated by clarity, placement, and national legal interpretation.
This reveals a tension between legal clarity and regulatory expectations for machineinterpretability required by European law. A possible solution is to recognize textual formats as “machine-readable” if they are standardized, accompanied by clear semantics, and placed within structured areas of the page. More supporting evidence from case law or clarifications in national laws is still needed.
Regardless of how many regulatory or technological questions arise, we observe an increase in the legal relevance of the copyright notice.
At the same time, only half of companies from our list actually use the phrase “All rights reserved.”
The longitudinal data from Table 4 provides additional insight into the evolving function of copyright notices within the EU context. Over time, we observe a clear shift toward UX-oriented and branding-driven patterns, for instance, increased consistency in the use of the © symbol, greater emphasis on visual minimalism (e.g., reduced mention of “Copyright”), and improved rights holder accuracy.
Paradoxically, however, the use of the phrase “All rights reserved” – perhaps, legally significant element under Directive 2019/790 – has not increased across the archived EU-based companies. Its frequency remains largely stagnant, despite its rising legal relevance in the context of TDM. This suggests a disconnect: the most doctrinally potent feature is the least responsive to regulatory change. Even more paradoxically, in the full global sample, the phrase “All rights reserved” appears more frequently than in the EU subsample. This could reflect a lag in regulatory awareness among European website operators/designers, or a broader dominance of design culture over legal incentives, or greater influence of the traditions (but not the regulations) of Pan-American law.
Notably, 89% of websites belonging to EU-headquartered companies from the sample used .com domains, which may limit the visibility of jurisdiction-specific adaptations and reduce the apparent influence of European legal frameworks on copyright notice design.
These findings could reinforce a central argument of this paper: copyright notices function as hybrid instruments. Their evolution is not linear or policy-driven; now, it is influenced by overlapping expectations from designers, marketers, legal departments, and users.
Nevertheless, we argue that textual notices such as “All rights reserved” may retain practical legal significance, especially when clearly worded, placed in a conventional legal location (e.g., the footer), and visible. Despite their long-standing presence, traditional copyright markers remain underutilized in current debates on machine-readable rights management. While researchers’ attention has shifted toward technically complex opt-out mechanisms like TDM flags, simpler textual opt-outs already embedded in digital content may provide a more immediate and effective path to enforceability. The simultaneous use of the © symbol, the word “Copyright,” and the phrase “All rights reserved” was identified in 18% of cases. This likely reflects legal uncertainty or a lack of understanding about the specific value of each element, prompting overuse to avoid potential mistakes.
Whether to include © / “Copyright” separately or together is not subject to any legal limitation. For example, under 17 U.S.C. § 401(d), even the use of word “copr.” is allowed. The majority of the websites studied use the © symbol. A significant share also adds the textual version, which is unjustified strictly from a UX perspective, since the © symbol is widely recognizable and selfsufficient. Moreover, the © symbol can be treated as a reliable anchor for machine identification, given its standardized use and distinctive visual form. In automated content extraction such markers are critical.
As shown by our research, limited attention is paid to the identification of the rights holder: in some cases, the notice failed to name any rights holder at all; in half the cases, the name does not match the legal name; or formats are used like “copyrights owned by the website” or “copyrights owned by the trademark.”
Such inconsistencies undermine the legal reliability of the notice. Economic copyrights must belong to a clearly identified natural or legal person – this is what gives the notice legal meaning. Since the study was based on websites belonging to legal entities, the information about their legal name and organizational form should have been included. Inaccurate or inconsistent naming may reflect a lack of coordination between design teams and legal departments, particularly in larger corporate structures where intellectual property may be managed by subsidiaries.
It is often the case (including among companies in our sample) that subsidiaries are created specifically to manage intellectual property, while the notices refer to the parent companies. This suggests a lack of coordination between designers and legal departments. The approach of several companies indicating their name options is useful – this enhances the long-term value of the notice.
We additionally examined the relationship between the use of the phrase “All rights reserved” and the accuracy of legal entity identification. The results differed only slightly from the overall sample: 47% of the websites using the phrase listed an incorrect legal name, of which 54% contained only minor inaccuracies.
This issue also exposes a deeper theoretical point: copyright notices exist at the intersection of legal precision and communicative simplicity. Designers and UX professionals often prioritize clarity and aesthetics, while legal practitioners require specificity and accuracy. This tension is visible in other elements as well – such as the indication of the year.
The question of indicating the year is also controversial, as the results show. A website contains
many works. From a legal standpoint, it would be accurate to specify the year the website was
created (for example, as 17 U.S.C. § 401(b)(2) suggests), but this may contradict the principles of
web design: as we have substantiated earlier, users might interpret it as a sign of an outdated, or
inactive, or even scam website [
Designers often favor displaying the current year to signal recency and administrative upkeep, even if it lacks direct legal precision. A date range, in turn, implicitly extends the scope of copyright claims over time – a stronger legal signal, but less visually concise.
The use of a date range expands the scope of the copyright claim: by using a range, the website signals that it continues to assert copyright over content created during those years, not just at the time of initial publication. However, it still remains unclear when each specific work on a website was created. This also creates text overload for the small notice.
Thus, the choice of how to present dates reflects both legal and behavioral considerations. An outdated year may undermine credibility, while a date range may suggest historical depth or continuous protection. Yet none of these conventions are standardized or legally mandated by consensus in the online environment, but in practice the current date prevails significantly. This reinforces the dual nature of the notice: it is simultaneously a legal assertion and a symbolic marker that communicates trustworthiness, professionalism, or legal awareness to users.
Some companies (often due to licensing agreements) include copyright attributions under every image, particularly when using stock photos. This may enhance legal traceability but creates visual clutter and inconsistency, especially if similar attributions are not provided for textual content. Such cases highlight a lack of holistic intellectual property approach at the design level.
In conclusion, the use of unilateral copyright notices is a complex and debatable practice. The presence of a copyright notice may act as a low-cost signal that reduces information asymmetry between content providers and users. Standardized copyright declarations can lower transaction costs by clarifying ownership, reducing the need for legal verification, and enabling smoother use in TDM contexts. Yet in the Digital Age, it gains new meaning and legal consequences – the rights holder can shape these outcomes positively or negatively, depending on whether the notice is formulated competently or poorly.
To structure our findings, we analyzed the empirical evidence in light of four main hypothetical functions of copyright notices: legal, behavioral, UX-related, and marketing-oriented. Table 5 below summarizes how each of these hypotheses aligns with observed indicators, the strength of supporting evidence, and the limitations encountered. This typology helps disentangle overlapping motivations behind the use of copyright notices and provides a clearer analytical framework for interpreting their role in the digital environment.
often missing or national interpretation incorrect regardless and formatting; lack of of the use of “All structured metadata; rights reserved”; machine-readability inconsistent years. rarely confirmed.
Some use of “All rights reserved”; includes legal names and years in notices; supports doctrinal intent; simultaneous use of multiple legal indicators suggests an attempt to maximize legal signaling or ensure compliance; © may be recognized by a machine; strengthens the evidence base in some jurisdictions.
Use of copyright symbols to Notices often vague deter unlawful copying; or redundant; presence of notices aligns lacking directive with user behavior theories. clarity; placement alone does not ensure behavior change.
prevalence of © symbol; consistent footer placement with other legal statements; efforts to avoid outdated
Marketing or Reputational years or visual clutter.
legal names; space used to highlight other IP to clients and partners; notices reinforce general perception.
branding logic; limited visual customization for brand.
Often not clearly motivated by marketing; may be incidental rather than intentional; only few explicit cases of IP portfolio signaling.
The legal and behavioral functions often reinforce each other – both rely on signaling deterrence, either through legal formalism or cognitive cues. Similarly, UX and marketing functions align through emphasis on design quality and credibility. However, tensions emerge between legal and UX aims: precise legal statements may clutter visual design. Likewise, behavioral clarity may conflict with marketing-driven ambiguity or branding flexibility. These intersections highlight the hybrid and sometimes contradictory nature of copyright notices in digital environments.
This study confirms that copyright notices continue to hold multifaceted value as a legal instrument, UX convention, behavioral cue, and branding tool. Their frequency of use reflects institutional and professional adherence rather than mere legal necessity.
The study also confirmed the behavioral and UX-related functions of the notice (albeit to a lesser degree than in previous research). We have discussed the new legal relevance of the notice, particularly within the EU context, especially in relation to Article 4 of Directive 2019/790 and the notion of “machine-readable” reservations. However, the effectiveness of plain-text notices remains uncertain, and judicial interpretations vary across Member States.
We recognize that some of the hypothetical functions (particularly the legal ones) require further empirical validation. The extent to which notices improve enforcement outcomes, such as in TDM contexts, remains to be proven. We avoid assuming causality where the evidence remains ambiguous. We also identified a separate marketing function, which lies in using this space to inform users about the brand and intellectual property rights, and to strengthen the reputation of a legally responsible company. From an economic perspective, copyright notices function as branding assets that support reputational capital.
Overall, the study contributes to bridging doctrinal copyright analysis with empirical web data, offering evidence that seemingly minor textual conventions play a measurable role in shaping AIera data governance.
Clarifying how textual copyright reservations are interpreted by AI systems is therefore critical, both to prevent unlawful scraping and to enable responsible TDM-based innovation in generative AI research. Future research should further examine how courts across different EU jurisdictions interpret the machine-readability requirement under Article 4 of Directive 2019/790, particularly in relation to plain-text notices. It would also be valuable to examine this issue in other jurisdictions, in light of technological and regulatory developments. In addition, future work could develop typologies of usage contexts and test which formulations most effectively fulfill each of these overlapping goals.
During the preparation of this work, the authors used Chat-GPT-5 in order to: Grammar and spelling check. [21] M. Chen, Improving website structure through reducing information overload, Decision
Support Systems 110 (2018) 84-94. doi:10.1016/j.dss.2018.03.009. [22] J. A. Mitchell, Putting some product into work-product: corporate lawyers learning from designers, Berkeley Business Law Journal 12(1) (2015) 1. [23] N. Allott, B. Shaer, Some linguistic properties of legal notices, Canadian Journal of Linguistics 58(1) (2013) 43-62. doi:10.1017/S0008413100002516. [24] D. Shmatkov, Intellectual Property Management of Industrial Software Products: The Case of Triol Corp., in: 2021 IEEE 8th International Conference on Problems of Infocommunications, Science and Technology (PIC S&T), IEEE, 2021, pp. 108-112. doi:10.1109/PICST54195.2021.9772237. [25] E. Goryunov, All Rights Reserved: Does Google's Image Search Infringe Vested Exclusive
Rights Granted under the Copyright Law, John Marshall Law Review 41(2) (2008) 487-526. [26] C. Sganga, M. Contardi, P. Turan, C. Signoretta, G. Bucaria, P. Mezei, I. Harkai, Copyright flexibilities: mapping and comparative assessment of EU and national sources, SSRN (2023).
Available at: https://ssrn.com/abstract=4325376. [27] W. A. Logan, J. Linford, Contracting for Fourth Amendment privacy online, Minnesota Law
Review 104 (2019) 101-170. [28] A. R. Puig, Reglamento de inteligencia artificial y derechos de autor, Revista Jurídica de les
Illes Balears 26 (2024) 45-72. doi:10.36151/RJIB.2024.26.02. [29] P. Mezei, A saviour or a dead end? Reservation of rights in the age of generative AI, European
Intellectual Property Review 46(7) (2024) 461-469. [30] J. L. O. Font, Derechos de autor y entrenamiento de sistemas de IA generativos: las obligaciones de transparencia y la minería de textos y datos en la normativa europea, IDP. Revista de Internet, Derecho y Política 42 (2025) 1-13. doi:10.7238/idp.v0i42.431327.