Literal translation is the most basic strategy, requiring the translator to preserve, as much as possible, the formal features of the source text while accurately conveying its content and form. This method is often used when the source and target languages share cultural and linguistic equivalences, as it retains the original imagery and structure ([9]).

Example 1: 我们要 坚定不移 贯彻“一国两制”“港人治港”“澳人治澳”、高度自治方针。

AI Version: We will unwaveringly implement the policies of “One Country, Two Systems”, “Hong Kong people administering Hong Kong”, “Macao people administering Macao”, and a high degree of autonomy.

Human Version: We will resolutely implement the policy of One Country, Two Systems, under which the people of Hong Kong administer Hong Kong and the people of Macao administer Macao, both with a high degree of autonomy ([4]).

The AI version translates 坚定不移 (jian ding bu yi) literally as “unwaveringly”, which, according to the Oxford Learner’s Dictionaries, means “in a way that does not change or become weaker in any way”, emphasizing steadfastness of belief or stance. However, from the perspective of the Skopos Rule, the emphasis here is on decisive action in implementing the policy, not on conviction. “Unwaveringly” falls short in conveying the executive force of policy implementation. The human translator opts for “resolutely”, which means “in a way that is strong and determined”. It focuses on decisiveness at the action level, more precisely aligning with the communicative purpose of demonstrating governance resolve.

Example 2: ……并与中长期发展目标相衔接，突出 迎难而上 、奋发有为的鲜明导向。

AI Version: These efforts have been aligned with our medium- and long-term development goals, highlighting a clear orientation of rising to challenges and striving to make progress.

Human Version: A target of around 5 percent is well aligned with our mid- and long-term development goals and underscores our resolve to meet difficulties head-on and strive hard to deliver ([3]).

迎难而上 (ying nan er shang) means to face difficulties without retreat and break through with determination. The AI translates it as “rising to challenges”, which suggests a reactive posture and weakens the sense of urgency and initiative. From the Skopos Rule, it inadequately conveys the government’s proactive determination in the face of economic pressures, thus reducing the effectiveness of external communication. The human translator’s “meet difficulties head-on” preserves the active and direct nature of the action, satisfying the Fidelity Rule while effectively serving the core purpose of conveying confidence in governance.

Example 3: 政府工作人员要 恪尽职守 、廉洁奉公，厉行法治、依法办事……

AI Version: Government personnel must perform their duties with dedication , act with integrity and serve the public interest, strictly abide by the rule of law and handle all matters in accordance with the law…

Human Version: All of us in government must perform our duties , be upright and devoted, respect the rule of law, and act in accordance with the law… ([3])

恪尽职守 (ke jin zhi shou) emphasizes that government functionaries should strictly fulfill their prescribed duties. The AI adds “with dedication” as a modifier. According to the Oxford Learner’s Dictionaries, “dedication” refers to “someone puts into an activity because they think it is important”, which introduces a subjective sense of devotion beyond the original meaning of strictly performing one’s assigned responsibilities, thus deviating from the Fidelity Rule. Moreover, the AI uses “their”, distancing government personnel from the reader, whereas the human translator’s “our” integrates the translator with the policy implementers, demonstrating greater rigor in fidelity.

Amplification refers to the strategy of supplementing information that is implicit in the source text but not explicitly stated, in order to make the translation more concise and fluent and more in line with target-language conventions, without compromising the core information. This technique is particularly useful when cultural gaps exist, as it adds necessary explanations or logical connectors to ensure coherence and readability for the target audience ([9]).

Example 4: ……我们统筹稳增长和增后劲，突出 固本培元 ，注重精准施策。

AI Version: …we have taken coordinated steps to stabilize growth and boost long-term vitality, with emphasis on consolidating fundamentals and adopting targeted policies.

Human Version: …we took coordinated steps to ensure stable growth and sustain its momentum, and gave priority to consolidating the foundations of the economy and taking targeted measures ([2]).

Since 固本培元 (gu ben pei yuan) originates from traditional Chinese medicine, a literal rendition would be abstract for foreign readers. The AI translates it as “consolidating fundamentals”, retaining the literal correspondence but failing to anchor it to the specific context. It results in obscure meaning, which not only fails to meet the semantic clarity required by the Coherence Rule for target readers, but also deviates from the communicative purpose of external publicity ([8]). The human translator adds “of the economy” after “foundations”, explicitly specifying the policy direction of consolidating the economic base. This not only removes the comprehension barrier caused by the culturally loaded term but also accurately conveys the source semantics under the Fidelity Rule, while meeting the contextual adaptation requirement of Skopos Theory.

Example 5: 回顾过去一年，多重困难挑战交织叠加，我国经济波浪式发展、曲折式前进，成绩 来之不易 。

AI Version: Looking back on the past year, multiple difficulties and challenges have intertwined and compounded. China’s economy has progressed in waves and advanced amid twists and turns. It has not been easy for us to deliver such achievements.

Human Version: Looking back at 2023, we can see that as we faced an array of interwoven difficulties and challenges, China’s economy grew in a wave-like fashion amid twists and turns. Indeed, our achievements did not come easily ([2]).

来之不易 (lai zhi bu yi) is a concluding evaluation of the economic development process. The AI restructures the sentence into an “it” expletive construction, but the transition from narration to evaluation feels abrupt, lacking coherence. The human translator adds the connective “Indeed”, reinforcing the logical link with the preceding narrative of twists and turns, and making the logic more explicit. Additionally, the human version supplies the subject “our achievements” and uses the idiomatic English negative structure “did not come easily”, resulting in overall coherence and fluency, thus enhancing the readability of the publicity-oriented text under the Skopos Rule.

Example 6: 千方百计 推动农业增效益、农村增活力、农民增收入。

AI Version: We will make every effort to improve agricultural efficiency, boost rural vitality, and increase farmers’ incomes.

Human Version: We will adopt all means necessary to raise agricultural returns, invigorate rural development, and increase rural incomes ([3]).

The AI translates 千方百计 (qian fang bai ji) as “make every effort”, a common everyday expression. It retains the idea of trying hard, but fails to convey the unsparing administrative determination behind China’s rural development work. From the Skopos Rule, it lacks the resolve and emotive force required for effective external communication. The human translator adds “necessary” after “all means” to reinforce the sufficiency and necessity of the measures, turning an abstract attitudinal expression into a concrete policy action. Moreover, the nominal structure “adopt all means necessary” better conforms to formal written English, improving the formality and communicative impact under the Coherence Rule.

Category shift, proposed by British linguist John C. Catford, specifically refers to shifts at the word-class level. Such shifts are common in English-Chinese translation. Through conversion from verbs to nouns, adverbs to prepositional phrases, etc., they break the constraints of source-text word classes, enhance textual coherence, and align with target-language expressive preferences ([1]). While not explicitly named in some functionalist taxonomies, this strategy falls under the broader principle of flexible adjustment to meet target-language norms, as advocated in Skopos-oriented translation practice ([9]).

Example 7: 严肃财经纪律，加强财会监督，严禁搞面子工程、形象工程，坚决制止 铺张浪费 。

AI Version: We will strictly enforce fiscal and financial discipline, strengthen accounting supervision, prohibit vanity projects and image projects, and resolutely stop extravagance and waste .

Human Version: We must tighten up financial and economic discipline and intensify oversight on accounting. Prestige and vanity projects and wasteful and excessive spending will be strictly prohibited ([2]).

铺张浪费 (pu zhang lang fei) is a verbal idiom emphasizing excessive and wasteful expenditure. The AI converts it into the nominal coordinate structure “extravagance and waste”, but the two abstract nouns lack concrete reference, making the criticism vague and lacking in clarity under the Coherence Rule. The human translator uses the adjective “wasteful and excessive” to modify the noun “spending”, transforming the Chinese dynamic criticism into an English static “adjective plus noun” evaluative structure. This preserves the negative stance while making the criticized object more concrete and perceptible, better satisfying both fidelity and coherence ([6]).

Example 8: ……保障人民群众 安居乐业 、社会安定有序。

AI Version: …so as to ensure that our people live and work in peace and society is stable and orderly.

Human Version: …so as to ensure that our people live and work in peace and contentment in a society of order and stability ([3]).

安居乐业 (an jv le ye) consists of two parallel verb phrases, where 安 (an) and 乐 (le) function as adverbs modifying 居 (jv) and 业 (ye) respectively. The AI renders it as “live and work in peace”, using a “verb + prepositional phrase” structure but omitting the component corresponding to “contentment”, thus failing to fully convey the meaning of taking joy in one’s occupation, which constitutes an information loss under the Fidelity Rule ([8]). The human translator retains both layers of meaning, converting it into “live and work in peace and contentment”, which completes the category shift from dynamic to “verb + prepositional phrase” while enhancing readability and achieving full semantic transfer under the Coherence Rule.

Example 9: 二是 坚定不移 全面深化改革扩大开放，增强发展内生动力。

AI Version: Second, we will remain committed to deepening reform and expanding opening up comprehensively, and boost the internal drivers of development.

Human Version: 2. We deepened reform and opening up comprehensively with unswerving commitment and boosted the internal momentum of development ([3]).

坚定不移 (jian ding bu yi) is a coordinate adverbial phrase functioning as a modifier in the source sentence. The AI and human translations differ in temporal framing: the AI maintains a future tense corresponding to the source text, while the human translation adopts a past-tense narrative, reflecting the retrospective structure of the Report on the Work of the Government ofChina. And in terms of strategy, the AI renders the idiom as the copula structure “remain committed to”, preserving adjectival quality but altering syntactic function. The overall reliance on the preposition “to” makes the sentence cumbersome, lacking fluency under the Coherence Rule. The human translator uses the preposition “with” plus the noun phrase “unswerving commitment” to form an adverbial phrase of manner, which conforms to the common English “verb + prepositional phrase” pattern, resulting in a compact and logically clear structure, thereby effectively enhancing fluency and acceptability under both the Coherence and Skopos Rules.

Overall, other strategies such as omission, restructuring, and retaining source-language imagery are also widely used in translating idioms in the Report on the Work of the Government ofChina, providing translators with flexible tools to serve the overall goal of external publicity and to comply with the three rules of Skopos Theory ([6]). For instance, omission is employed when redundant modifiers are deleted to achieve conciseness, and restructuring is used to reorder sentence elements to match English hypotactic logic.

There are also cases where AI and human translations produce identical results, such as 稳中求进 (wen zhong qiu jin) rendered as “pursue progress while ensuring stability”, 坚如磐石 (jian ru pan shi) as “rock-solid”, and 自立自强 (zi li zi qiang) as “self-reliance and strength”. These idioms are relatively straightforward in meaning and have fixed or highly corresponding expressions in English. Therefore, one should neither dismiss AI’s capabilities outright nor adopt its outputs uncritically. Instead, translation tasks should be differentiated by type and difficulty to explore more targeted pathways for human-machine collaboration.

As can be seen from the above cases, the discrepancies between AI and human translations are not accidental but result from differing translation intentions and practical logics.

First, human translators prioritize purpose, whereas AI leans toward lexical and phrasal matching. Vermeer’s Skopos Theory posits that translational action is determined by its intended purpose. AI translation differs significantly from human translators in handling deep meaning, metaphor, and cultural background ([8]). Human translators, upon reading the source text, first judge the function of the idiom or sentence, and then adopt corresponding strategies. For instance, 坚定不移 (jian ding bu yi) is rendered as “resolutely” to highlight decisiveness in policy implementation. AI, by contrast, tends to produce outputs that reflect high-frequency collocations from its training data, as observed in the generated translations ([10]). It outputs “unwaveringly”, which leans toward steadfast belief, deviating from the context of policy execution.

Second, human translators emphasize overall structural balance, while AI tends to obsess over minor details. AI translation takes the sentence as its unit, focusing on formal and literal correspondence between target and source, and neglecting deeper semantic or pragmatic analysis, which may lack individuality and creativity ([10]). For example, AI adds “with dedication” to 恪尽职守 (ke jin zhi shou), introducing a sense of devotion beyond the original meaning of fulfilling duties, but violating the Fidelity Rule. Human translators usually retain only the core meaning without arbitrary additions.

Third, human translators are more sensitive to textual coherence, whereas AI tends to produce mechanical restructuring. Human translators are adept at constructing causal and intentional connections, making the underlying logical relations more explicit. While AI shows clear deficiencies in deep cohesive ties ([10]). Most Chinese idioms appear in parallel or antithetical forms. So, when translating into English, it is often necessary to adjust word order and convert forms to clarify implicit logical relations. For instance, when 坚定不移 (jian ding bu yi) functions as an adverbial idiom, AI uses the copula structure “remain committed to”, making the sentence loose and verbose due to reliance on the preposition “to”. The human translator, however, uses the prepositional phrase “with unswerving commitment” as an adverbial of manner, highlighting the modifying relationship between sentence components.

Based on the above analysis of the causes of differences, it is reductive to simply conclude that “human translation is superior” or that “AI will replace human translation”. Rather, it is necessary to recognize the respective strengths of both and explore a complementary and cooperative model of coexistence. AI and human translation are not in a contradictory, zero-sum relationship but are mutually reinforcing and complementary.

First, position AI as a provider of detailed initial drafts. “AI translation + post-editing” has become a mainstream working mode in the translation industry ([10]). However, machine translation still needs improvement in phonological harmony, refined wording, and rhetorical appropriateness. For idioms with relatively straightforward meanings and low cultural loading, AI can produce direct translations that need only minor adjustments by human editors. For semantically complex idioms, AI’s literal versions, though not necessarily concise, can often present each semantic component in detail, allowing human translators to clearly see what details AI has preserved and decide what to omit or convert.

Second, position human translators as purpose-oriented reviewers and refiners. In the AI era, the translator’s role is shifting from traditional text converter to editor, reviser, and quality control expert. Human translators should first apply the Skopos Rule to check whether the translation serves the core purpose of external communication, then apply the Coherence Rule to judge whether the English is fluent and logically clear, and finally apply the Fidelity Rule to confirm semantic accuracy. The construction of the translator’s subjective agency helps enhance translation competence, which is precisely the irreplaceable value of human translators in human-machine collaboration ([1]).

Third, build a dedicated database of “purpose + strategy”. The 38 idioms extracted from the Reports on the Work of the Government ofChina (2024-2026) in this study, along with their optimal translations under various contexts, can serve as training data for AI. By enabling AI to learn when human translators use omission, amplification, category shift, and other strategies, the model can gradually acquire purpose-oriented strategic decision-making while retaining its strengths in detail processing. Accumulating a translation database of “purpose + strategy” to help AI progressively develop contextual awareness of strategy selection represents an important direction for translation studies in the AI era.

In summary, in the human-machine collaborative translation model, AI is better suited to act as a meticulous initial translator and a checker of semantic expression, while human translators serve as gatekeepers of translation purpose, determining the stylistic tone. Each has its strengths. Therefore, the key is to find appropriate divisions of labor so that AI’s efficiency and human precision form a synergy to jointly enhance translation quality and efficiency.