Inclusion criteria

It was decided to include empirical studies that met the following criteria:

Studies that did not report implemented and evaluated interventions were excluded, such as exploratory, descriptive, correlational, or confirmatory research that assessed reading comprehension performance using digital technologies but without implementing an instructional program. Studies conducted outside Ibero-America, aimed at speakers of another language or Spanish as a second language (L2), that pursued the improvement of other reading skills than comprehension; interventions aimed at early childhood education students, higher education students, or students without typical development, as well as studies in which the intervention was not exhaustively described, were also excluded. In addition, conference presentations, conference papers, books, undergraduate, masters, and doctoral theses, among others, were excluded.

The search was limited to the last ten years in order to focus on studies that incorporate recent digital technologies, given that substantial changes have occurred in the last decade both in available technologies and in instructional designs that integrate technologies in educational contexts.

Search procedure

The systematic review was conducted in two stages. The main search was carried out during the first week of December 2023, while in September 2025 an update search was performed. On this second occasion, a temporal filter was applied across all databases, restricting the search to the 2023-2025 period. In addition, a further inclusion criterion was applied: (8) the study had not been included in the previous search.

Six databases were selected: Scopus, Web of Science, ERIC, SciELO, LA Referencia, and Dialnet. The search was carried out based on combinations of keywords, using AND and OR operators, organized into three conceptual axes: one related to reading comprehension, another to intervention, and a third linked to digital technologies (Table 1).

Terms for this last axis were selected by conducting a screening of 27 articles related to the topic in order to include as many technology-related terms as possible. A temporal filter was applied to the initial search across all databases. In La Referencia, the first search string was restricted to title field only, whereas in Dialnet a reduced search string was used due to the characteristics of the database. In Scopus and Web of Science, searches were conducted in English using translated versions of the keywords, and results were filtered by country. In ERIC, the search was conducted using thesaurus-based descriptors.

Table 1: Search terms and operators used in each database 

Once the search was completed, duplicates were detected and an initial selection was made based on title and abstract using Rayyan software (^{Ouzzani et al., 2016}). To this end, all records were screened by the first author; the second author independently screened a randomly assigned 50 % of the records, and the third author screened the remaining 50 %. Consequently, each record was assessed for inclusion by two reviewers. This process was carried out independently by each author, yielding an agreement percentage close to 95 %, with a strong agreement (k = .68). Disagreements were resolved through dialogue. After this initial selection, full texts were read.

Analysis of studies

After the final reading and selection of publications that met inclusion criteria, an analysis of the characteristics of intervention proposals was conducted in order to produce a qualitative synthesis (^{Moher et al., 2009}). To this end, a data matrix was constructed in Excel in which, in addition to bibliometric variables of the articles, the following aspects were examined in line with the proposed objectives: methodological design, sample size, main results, type of technology employed, cognitive and metacognitive processes addressed, structure of the sessions, feedback provided during the sessions, and impact of the interventions on reading comprehension.

General characteristics of intervention proposals

Studies were mostly conducted with populations from Spain (n = 8), followed by Chile (n = 3), Colombia (n = 2), Ecuador (n = 2), and Peru (n = 1) (Table 2). Eleven studies were implemented in primary education, four in secondary education, and one in both educational levels.

Across the included studies, there was a tendency toward the use of a quantitative methodological approach with pre- and post-test designs. Seven studies employed a design with one experimental group and one control group (^{Benavides & Zambrano, 2023}; ^{García et al., 2021}; ^{Quiroz-Benítez et al., 2023}; ^{Salmerón & Llorens, 2019}; ^{Serrano et al., 2018}; ^{Serrano-Mendizábal et al., 2023}; ^{Vidal-Abarca et al., 2014}); in two studies, a design with two experimental groups and one control group was used (^{Guerra & Mellado, 2017}; ^{Torres & Medina, 2020}); and five studies employed a single-group design (^{McCarthy et al., 2020}; ^{Salmerón et al., 2015}; ^{Sánchez & Pascual, 2022}; ^{Serrano et al., 2018}; ^{Soto et al., 2019}). One study adopted a qualitative approach (^{Pedrozo Castrillo et al., 2021}), and another employed a mixed-methods approach, a design-based research with one experimental group and one control group (^{Tobar-Muñoz et al., 2017}).

The interventions reported in the studies varied considerably in terms of total duration and frequency of sessions, ranging from 1 to 50 sessions, each lasting between 15 and 50 minutes (Table 2 . . .).

Table 2: Characteristics of the included studies in the systematic review 

^Note: In the Educational level column, the denominations reported in each study were used, according to the educational system of each country. In the Design and sample column, the designs are reported as described in the original studies. When available, participants’ age and/or age range were included. A distinction is made between control group (CG) and active control group (Active CG) in order to account for groups in which participants engage in activities, receive sessions, or follow a program with similar tasks to the implemented training in the experimental group (EG)

Type of digital technology used

All studies used technologies designed for educational purposes, which were mainly implemented by using computers.

Many of the studies (n= 5) used ITS: I-Start-E (^{McCarthy et al., 2020}), TuinLEC (^{Vidal-Abarca et al., 2014}), TuinLECweb (^{Serrano et al., 2018}; ^{Serrano-Mendizábal et al., 2023}), and AutoLEC (^{Serrano-Mendizábal et al., 2023}). ITS are interactive learning environments that use artificial intelligence to adapt to students' skills (^{Xu et al., 2019}). Based on system-student interactions, the system assesses their knowledge, offers feedback, and creates instruction tailored to their performance.

Three studies used video modeling or Eye-movement modeling examples (EMME) (^{García et al., 2021}; ^{Salmerón et al., 2015}, Salmerón & Llorens, 2019). This technology consists of an instructional technique that provides a model of a subject's actions related to a given task. It not only allows strategies employed by the subject to be modeled, but also their visual processing and attention to the task (^{van Gog et al., 2024}). Studies that use this technology for reading digital texts emphasize modeling the subject's strategic processing of the information that needs to be prioritized in order to understand the texts. In the study by ^{García et al. (2021)}, this technology was combined with software for creating concept maps (CmapTools), in line with the aim of applying knowledge derived from the models to construct an explanatory map on a given topic, with an appropriate hierarchical organization of information.

Four studies used websites for their development, in some cases complemented by other digital tools. ^{Soto et al. (2019}) employed COMPRENDE, an educational technology that focuses on reading strategies and combines presentation of lessons, activities, and feedback provided by an animated educational agent. Meanwhile, the study by ^{Guerra and Mellado (2017}) used an assisted-reading book or A-book. This book is distinctive in that it presents a selection of stories and related questions in digital format, and supports readers through feedback provided during reading. ^{Benavides and Zambrano (2023}) developed an intervention through an educational environment supported by a task selection algorithm, designed to promote self-regulation in reading tasks by adjusting activities according to the relationship between mental effort and performance, as reported by each student. To explain how they should rate effort and select tasks, they included video modeling, and then students had to complete reading activities through Google Forms, assigned based on an algorithm that took into account complexity level and support degree (no support, low, or high). ^{Quiroz Benítez et al. (2023}) implemented a proposal using Facebook as an educational environment for sharing materials, producing content, and carrying out collaborative activities linked to certain reading tasks, established in advance and based on certain cognitive processes.

Two studies used Virtual Learning Objects (VLO; ^{Pedrozo Castrillo et al., 2021}; ^{Torres & Medina, 2020}). VLO are characterized as digital resources designed and structured for an educational purpose, which present specific content to students (^{Cóndor-Herrera et al., 2021}). Pedrozo Castrillo et al. (2021) used a platform (educaplay) that allows the design of activities, with the aim of working on reading comprehension. ^{Torres and Medina (2020)} employed virtual classrooms on a Learning Management System (LMS) platform. In the classroom, VLOs were presented differently depending on the experimental condition (passive vs limited interactivity).

Finally, two studies adopted a game-based learning framework (^{Sánchez & Pascual, 2022}; ^{Tobar-Muñoz et al., 2017}). ^{Sánchez and Pascual (2022)} implemented a digital serious game developed on an educational web platform (supertics). Unlike “conventional” games, serious games are designed with the primary goal of serving as educational tools, taking advantage of the affordances and benefits offered by games (e.g., their appeal in terms of fun) while subordinating these features to a secondary role (^{Calvo-Ferrer, 2018}). In turn, ^{Tobar-Muñoz et al. (2017)} adopted a novel approach by combining an augmented reality digital game with storybook reading. In this way, the augmented reality game unfolded as the images accompanying the pages of the book were viewed using a tablet.

Cognitive and metacognitive aspects of reading comprehension

Except for ^{Quiroz-Benítez et al. (2023}), ^{Sánchez and Pascual (2022}), and ^{Pedrozo-Castrillo et al. (2021}), most studies aimed to address both cognitive and metacognitive aspects related to reading comprehension, even when the focus was primarily on one dimension or the other (Figure 2).

Figure 2:  

In the reviewed studies, the most frequent tasks were those related to metacognitive processes, which require assessing or monitoring one's own comprehension (^{García et al., 2021}; ^{Guerra & Mellado, 2017}; ^{McCarthy et al., 2020}; ^{Salmerón et al., 2015}; ^{Salmerón & Llorens, 2019}; ^{Serrano et al., 2018}; ^{Serrano-Mendizábal et al., 2023}; ^{Soto et al., 2019}; ^{Torres & Medina, 2020}; ^{Vidal-Abarca et al., 2014}). The second most frequent tasks were cognitive activities linked to hierarchization of textual information, mainly related to the identification of main ideas (^{Benavides & Zambrano, 2023}; ^{McCarthy et al., 2020}; ^{Quiroz-Benítez et al., 2023}; ^{Sánchez & Pascual, 2022}), the use of graphic organizers (^{García et al., 2021}; ^{Quiroz-Benítez et al., 2023}), or the preparation of summaries of the read texts (^{Soto et al., 2019}). Thirdly, there were cognitive activities that promote generation of inferences to connect two pieces of information or link the information presented in the text with prior knowledge (^{Benavides & Zambrano, 2023}; ^{Guerra & Mellado, 2017}; ^{McCarthy et al., 2020}; ^{Pedrozo Castrillo et al., 2021}; ^{Quiroz-Benítez et al., 2023}; ^{Sánchez & Pascual, 2022}; ^{Soto et al., 2019}; ^{Tobar-Muñoz et al., 2017}; ^{Torres & Medina, 2020}). Fourth, tasks related to self-regulation were proposed (^{Benavides & Zambrano, 2023}; ^{García et al., 2021}; ^{Salmerón & Llorens, 2019}; ^{Serrano et al., 2018}; ^{Serrano-Mendizábal et al., 2023}; ^{Vidal-Abarca et al., 2014}) as a metacognitive activity that is not only key to strategic reading-particularly of texts on screens- but also constitutes a fundamental component of autonomous learning, which is positively valued in virtual environments, where students are required to take on a more active role in regulating their own performance.

To a lesser extent, metacognitive activities linked to the formulation of hypotheses or predictions for reading (^{Benavides & Zambrano, 2023}; ^{Salmerón et al., 2015}; ^{Soto et al., 2019}; ^{Torres & Medina, 2020}), establishment of reading goals (Salmerón et al., 2015; Salmerón & Llorens, 2019; ^{Serrano et al., 2018}; ^{Vidal-Abarca et al., 2014}), and development of self-explanations (^{García et al., 2021}; ^{McCarthy et al., 2020}) are also reported.

Structure of the sessions

Most intervention proposals (n=10) were implemented as complementary spaces to regular lessons, without the intention of being directly integrated into curricular content. In addition, they were organized following a defined structure established prior to implementation. Thirteen proposals implemented an individual work format, where each student works with their own device. To a lesser extent, some proposals (^{García et al., 2021}; ^{Pedrozo Castrillo et al., 2021}; ^{Salmerón et al., 2015}) employed pair work, emphasizing the potential of interaction and discussion among students for learning.

A large number of studies adopted a phase-based structure (^{McCarthy et al. 2020}; ^{Salmerón & Llorens, 2019}; ^{Salmerón et al., 2015}; ^{Serrano et al., 2018}; ^{Serrano-Mendizábal et al., 2023}; ^{Soto et al., 2019}; ^{Vidal-Abarca et al., 2014}) or blocks-based structure (^{García et al., 2021}; ^{Guerra & Mellado, 2017}), alternating moments of instruction or teaching with moments of practice or training. In these cases, the initial phase or block was dedicated to explicitly introducing a certain process or skill related to reading comprehension, based on a resource available in the program or an explanation by the researcher or teacher. This introduction could be presented with or without other multimodal elements (e.g., images, graphics, or videos). Students were then presented with a practice or training section, focused on working on strategies or skills explained in the previous instance. In the study by ^{McCarthy et al. (2020)}, two sections were enabled: guided practice and open practice environment. In addition to the instruction and training phases, some studies (^{Salmerón et al., 2015}; ^{Salmerón & Llorens, 2019}) included a third phase of reflection on the content presented in the previous phases and/or on the students' responses to the tasks, with the aim of encouraging students to evaluate their own performance and identify areas for improvement.

^{Benavides and Zambrano (2023}) proposed an introductory stage, but aimed at explaining the sequence of the intervention proposal rather than teaching a specific strategy or skill. In this study, the intervention structure depends on the students' performance, as the tasks they must solve are assigned using an algorithm based on their assessment of mental effort. This results in the resolution of reading tasks with different levels of complexity and support.

In other studies, sessions were organized according to reading-related moments: before, during, and after reading (^{Pedrozo Castrillo et al., 2021}; ^{Torres & Medina, 2020}). In these cases, prior to reading, anticipation of the text's content and activation of prior knowledge were encouraged. Time was then devoted to reading the text, and finally, after reading, activities related to comprehension were carried out.

Some studies organized the intervention sequence around a game-based dynamic. ^{Sánchez and Pascual (2022}) used a game with scheduled sessions, in which a text was read and various related activities were carried out; whereas in the intervention proposed by ^{Tobar-Muñoz et al. (2017}), reading serves as an introduction to a challenge or goal that the student must solve, either through an in-game action or by answering to a question.

Finally, in the study by ^{Quiroz-Benítez et al. (2023}), the intervention proposal was organized according to reading tasks to be completed, which, in turn, aimed to promote a specific strategy to which students had to respond on a social network, sometimes individually and sometimes collaboratively.

Feedback provided during the sessions

In nine of the studies analyzed, feedback is an important element of the intervention, although its implementation varies according to the design of each proposal.

In the study by ^{Guerra and Mellado (2017}), feedback varied depending on the experimental condition: experimental groups received explanatory feedback, in which, regardless of each participant's performance, an explanation of the correct answer was provided, indicating the relevant and accurate information. In contrast, the control group was only informed whether the answer was correct or incorrect.

In some studies, students received feedback at specific phases of the intervention. For example, ^{Salmerón and Llorens (2019}) and ^{Salmerón et al. (2015)} introduced feedback during the reflection phase. Information was provided about strategies displayed in the EMMEs (experimental group) or in the written cases (control group) in order for students to reflect on their performance.

Meanwhile, in the studies by ^{McCarthy et al. (2020}), ^{Serrano et al. (2018}), ^{Serrano-Mendizabal et al. (2023}), and ^{Vidal-Abarca et al. (2014}), feedback was used at training phase and appeared after students had answered the question. In the study by McCarthy et al. (2020), the system assigned a score to students’ responses (from 0 to 3 points) and provided information on how to improve self-explanation. In the other three studies, feedback was presented in different ways. First, the correct answer was indicated; then, the information relevant to the question was highlighted in the text; and finally, students received adapted written feedback based on time spent, strategies used, and support tools used. ^{Serrano-Mendizabal et al. (2023)} also used another program (AutoLEC) which, although similar to TuinLEC Web format, differed in terms of the feedback provided, as it only indicated whether the student's answer was correct or not.

Finally, in the study by ^{Soto et al. (2019}), feedback was provided with a metacognitive purpose: to encourage students to detect errors and adjust and calibrate their reading. Thus, during the training phase, explanatory feedback was used with the purpose of encouraging students to reflect on their performance. The feedback was aimed at identifying whether the content used was the most appropriate and also at identifying inconsistencies in some texts.

Effects of the interventions on reading comprehension

In order to assess reading comprehension performance, eight of the studies used standardized measures, such as Lectum (^{Riffo et al., 2011}; used in ^{McCarthy et al., 2020}; ^{Soto et al., 2019}), Comprehension Strategies Test (^{Vidal-Abarca et al., 2007}; used in ^{García et al., 2021}; ^{Vidal-Abarca et al., 2014}), Comprehension Processes Test (^{Martínez et al., 2008}; used in ^{Serrano et al., 2018}), ECOMPLEC (^{León et al., 2012}; used in ^{Serrano-Mendizábal et al., 2023}) or the Reading Efficiency Test (^{Salmerón et al., 2015}). In addition, instruments specifically designed for the studies were employed, generally adopting a format consisting of reading a text and answering questions about it (^{Benavides & Zambrano, 2023}; ^{García et al., 2021}; ^{Guerra & Mellado, 2017}; ^{Pedrozo Castrillo et al., 2021}; ^{Salmerón & Llorens, 2019}; ^{Serrano et al., 2018}; ^{Tobar-Muñoz et al., 2017}; ^{Torres & Medina, 2020}) or instruments used in similar studies to measure comprehension levels (^{Quiroz-Benítez et al., 2023}). In some cases, performance was evaluated by calculating the scores obtained in an activity carried out through the technology used (^{Guerra & Mellado, 2017}; ^{Sánchez & Pascual, 2022}; ^{Serrano et al., 2018}; ^{Serrano- Mendizábal et al., 2023}). In addition, some studies incorporated other measures to analyze dimensions such as motivation (^{McCarthy et al., 2020}; ^{Tobar-Muñoz et al., 2017}) or students’ perceptions (^{McCarthy et al., 2020}; ^{Soto et al., 2019}; ^{Vidal-Abarca et al., 2014}).

Regarding the assessment moment, post-test was administered immediately after the intervention. Only two studies included follow-up measures to assess the stability of the effects two weeks after the intervention (^{Serrano et al., 2018}).

As shown in the results synthesis in Table 2, improvements in reading comprehension performance were generally reported (^{McCarthy et al., 2020}; ^{Pedrozo Castrillo et al., 2021}; ^{Quiroz-Benítez et al., 2023}; ^{Salmerón et al., 2015}; Salmerón & Llorens, 2019; ^{Sánchez & Pascual, 2022}; ^{Serrano et al., 2018}; ^{Serrano-Mendizabal et al., 2023}; ^{Torres & Medina, 2020}; ^{Vidal-Abarca et al., 2014}). Some studies reported specific improvements: for example, in the case of ^{Serrano et al. (2018)}, improvements in comprehension were greater for low-performing students. ^{Torres and Medina (2020)} reported a significant improvement in both experimental groups compared to the control group, although there were no significant differences based on the experimental condition (passive and limited interactivity). However, some studies reported specific aspects in which improvements were not significant after the intervention, such as integrated comprehension (^{García et al., 2021}), the strategy of activating prior knowledge (^{Salmerón et al., 2015}), or school grades (^{Sánchez & Pascual, 2022}). Finally, certain studies found no significant improvements in students’ reading comprehension (^{Benavides & Zambrano, 2023}; ^{Guerra & Mellado, 2017}; ^{Tobar-Muñoz et al., 2017}).