Ever since online psychotherapy was conceived as an alternative to ordinary therapy, especially during lockdown, there has been an evident increase in demand for online psychotherapy.

As Bauman (2005) introduced, we live in a liquid society which means that we are constantly adapting to the new stimuli we face. In the same liquid way, we can either choose to adapt and move forward or remain in conventional patterns. The most recent event occurring in our society to which we had to adapt is the digital revolution. Modern day lives revolve around electronic devices, so naturally the concept of online therapy comes as no surprise to the new generations. There are currently four generations living in the same decade, so their opinions tend to differ when it comes to the question if online psychotherapy offers the same depth as a face-to-face method. In a recent study, Yılmaz et al. (2024) found that opinions about online psychotherapy differ between young and middle-aged patients, with young people being more inclined towards online psychotherapy.

The current society is ruled by a harmonious fusion between immediacy and comfort.

It is also interesting to talk about how relationships have also changed in recent years. In a liquid society there are, needless to say, liquid relationships. We have come to a place where relationships are shaped by the values of our current society: immediacy and comfort. We cannot define modern day relationships without the concept of “digital.” This digital component has brought us closer in some aspects and more distant in others. People expect to connect with therapists in the same way as they do in personal relationships, introducing a screen between the two people which, unironically, can seem to bring them closer (Mercadal and Cabré, 2021).

Culture of immediacy is rapidly overtaking our society and especially the newest generations, who prefer digital over physical. It is easy to assume that they will also be more inclined towards online psychotherapy, which can explain the higher demands in online psychotherapy nowadays.

Online psychotherapy has been considered a complement to traditional face-to-face psychotherapy (Berle et al., 2015; Cipolletta et al., 2018; Norwood et al., 2018), but it has been shown to be a valid alternative in a number of studies when face-to-face psychotherapy is not possible (Cook and Doyle, 2002; Knaevelsrud and Maercker, 2006; Reynolds et al., 2006; Preschl et al., 2011; Wagner et al., 2014; Holmes and Foster, 2012; Lewis et al., 2021).

Since the late 1990s, research into online psychotherapy has focused on its effectiveness rather than knowing in which situations it might be helpful (Stoll et al., 2020). However, it was found that online therapy can be helpful in people who have difficulty asking for help, as it can provide a sense of shelter and protection (Schultze, 2006; Tate and Zabinski, 2004; Vallejo and Jordán, 2007), unless there is a hidden reason behind the client’s proposal to start online therapy, such as keeping distance or feeling more protected, because behind this request there could be relational problems that prevent them from coming (Cabré and Mercadal, 2016). Nevertheless, it is not recommended for patients with a lack of emotional control, such as psychosis, major depression, violence or in crisis situations (Torre and Pardo, 2019), as dealing with crisis and suicidality could be challenging in online psychotherapy. However, in patients diagnosed with depression and anxiety, combining face to face with online therapy was found to be more effective than just face to face therapy (Zwerenz et al., 2017).

Bibliography differs from opinions as some authors think that it is the therapist’s style and personality that determine the therapeutic relationship (Cipolletta et al., 2018) and therapy effectiveness (Rathenau et al., 2021) rather than the modality of the psychotherapy. As for outcomes in different psychotherapeutic models, Rosenzweig (1936) introduced the idea that all therapies work through common factors, meaning that different therapies produce similar outcomes (Budd and Hughes, 2009; Luborsky et al., 2002; Marcus et al., 2014; Cuijpers et al., 2018). Building on this concept, Lambert (1992) pie chart model and the study by Cuijpers et al. (2012), suggested that approximately 30% of therapeutic change can be attributed to common factors, 15% to specific techniques, 40% to additional therapeutic factors and 15% to the placebo effect. Key common factors include therapeutic alliance, therapist empathy, and client expectations (Browne et al., 2021). Horvath et al. (2011) proved that stronger therapeutic alliances are associated with better treatment outcomes. However, the evidence is correlational and cannot establish causality (Norcross and Wampold, 2011).

While therapeutic alliance is a common factor, it also plays a different role in each psychotherapy model. In therapies which emphasize therapeutic alliance, such as psychodynamic therapy, the alliance may be more linked to the outcome than in therapies that do not emphasize the therapeutic alliance, such as Cognitive Behavioral Therapy (CBT) (Huibers and Cuijpers, 2015; Romero-Moreno et al., 2023). Nevertheless, online CBT has been found to be the most effective, as it is also the most studied one (Weightman, 2020). Although, more research is needed to determine the online effectiveness of other types of psychotherapy such as psychodynamic (Carrillo et al., 2021; de Bitencourt Machado et al., 2016; McDonald et al., 2020).

Despite the proven relevance of common factors for effective psychotherapy, especially those of an emotional nature, Romero-Moreno et al. (2021) found that therapists still have doubts about the factors responsible for psychotherapeutic effectiveness. They do not place particular value on specific factors, such as the techniques used or the therapeutic approach adopted.

According to Romero-Moreno et al. (2021, 2023), therapists’ perspectives on the importance of emotional factors, including the therapeutic alliance, are shaped by their theoretical approach and their views on the relative effectiveness of different psychotherapy models. Hence psychodynamic therapists place greater importance on the therapeutic alliance, cognitive behavioral therapists emphasize directiveness and support, and eclectic therapists highlight common and emotional factors for therapeutic change.

The benefits of online psychotherapy found in the recent bibliography are that online sessions can reduce therapist’s workload and improve access to healthcare, particularly in rural areas (Stoll and Trachsek, 2019) whilst maintaining the quality of care (Carrillo et al., 2021). It can provide flexibility for both therapists and patients as it can be delivered in remote locations (Lippke et al., 2021).

Some of the challenges of online psychotherapy are actually reflected in the setting, as clients may be in the same environment as the people with whom they are struggling. Online therapy can overly condition the intimacy of the session and posteriorly the development of the treatment (Mercadal and Cabré, 2021). Another aspect altered in the setting is that online psychotherapy lacks a “frame” (Ashraf et al., 2020), which can provide a secure base and a reassuring effect whereby ‘even chaos, once it has been given a name, is less chaotic’ (Milner, 1950). The travel time also constitutes part of the setting, and it can give both patients and therapists the opportunity to reflect about the content of these appointments (Sayers, 2021).

Overall, these issues may explain the high dropout rates in online psychotherapy (Boldrini et al., 2020). Some other factors that should be addressed and may contribute to dropout are women, younger patients, unpartnered patients, less educated patients, patients with depressive symptoms and those with fewer expectations of therapy (Lippke et al., 2021).

Videoconferencing is an interaction with a great amount of sensory components (Cabré and Mercadal, 2016), and a common experience amongst therapists is that they feel a greater fatigue experienced with online activity (Mercadal and Cabré, 2021), also known as “zoom fatigue” due to excessive use of videoconferencing platforms, especially during the pandemic (Bennett et al., 2021; Epstein, 2020). In addition, therapists reported feeling a sense of loss of presence and engagement in the online mode (Gullo et al., 2022). Results consistent with Malouin-Lachance et al. (2025), who concluded that, while digital interventions improve accessibility and engagement in mental healthcare, they also present challenges related to limited emotional depth, personalization, and ethical considerations.

Yılmaz et al. (2024) used qualitative research to assess how patients perceived both methods. The most frequent metaphors about online psychotherapy were associated with the categories of convenience, artificiality, similarity to face-to-face psychotherapy, and ineffectiveness. And the metaphors about face-to-face psychotherapy belonged to the categories of contact, effectiveness, reality, and difficulty. Recently, Leuchtenberg et al. (2022) found that patients and psychotherapists mainly preferred face to face psychotherapy.

Ierardi et al. (2022) studied the “Effectiveness of an online versus face-to-face psychodynamic counseling intervention for university students before and during the COVID-19 period.” The aim of the study is to evaluate the effectiveness of the online counseling intervention during the COVID-19 pandemic, compared to face-to-face interventions. The results show that the online counseling intervention during the pandemic was effective in reducing psychological distress scales such as depression (p = 0.008), obsessive-compulsive (p = 0.008), interpersonal sensitivity (p = 0.005), and anxiety (p = 0.011), as well as the total scale of the SCL-90 R (p = 0.017). The face-to-face counseling intervention was effective in reducing psychological distress in all subscales and in the total scale of the SCL-90 R (p = 0.000) and in increasing the level of life satisfaction (p = 0.023). Attachment style did not influence the effectiveness of the online and face-to-face interventions. However, face-to-face interventions were more effective than online therapy in reducing a wider spectrum of psychopathological problems and in increasing life satisfaction.

To date, there is a gap in literature on the effectiveness of online psychotherapy compared to face-to-face psychotherapy. So far, it has been found to be effective (Barker and Barker, 2022; García et al., 2022), although online therapy does not provide any additional benefit when compared with to face to face therapy (Rollman et al., 2018). However, face-to-face interventions have proven to be more effective than online therapy in reducing a wider spectrum of psychopathological problems and in increasing life satisfaction (Ierardi et al., 2022).

As there is still some uncertainty about the impact of online therapy, before simply opting for an online format, Carrillo et al. (2021) suggest that we should consider which patients are most likely to benefit from a face-to-face format. Although online therapy can be delivered to almost all patients, not all patients are suitable for this modality (Cabré and Mercadal, 2016).

Essential aspects of psychotherapy are the therapeutic relationship and alliance, consistently ranked among the most reliable predictors of positive therapeutic change. They are regarded as a one of the most consistent predictors of effective therapeutic change and constitute a central focus across all major psychotherapy models (Norcross and Lambert, 2011, 2019; Flückiger et al., 2018; Wampold and Flückiger, 2023; Karunarathna and Jayawardana, 2024; Opland and Torrico, 2024; Malouin-Lachance et al., 2025). This is further supported by evidence indicating stronger alliances, characterized by a solid bond between therapist and patient, as well as agreement on therapeutic goals and tasks, have been found to be associated with better patient outcomes (Horvath et al., 2011; Wampold, 2015).

In recent studies, patients rated the therapeutic alliance higher than psychotherapists, a result consistent with previous studies (Lopez et al., 2019; Simpson and Reid, 2014), and therapists rated the therapeutic success nearly the same as patients (Eichenberg et al., 2022). Another study found out that while patients reported equivalence in face-to-face and online psychotherapy, therapists experienced an equivalent bond but more advantages in tasks and goals in face-to-face therapy (Leuchtenberg et al., 2022).

Face-to-face human treatment is not comparable to online therapy, although sometimes it can be a good resource, in no case it is more real than face-to-face interactions (Mercadal and Cabré, 2022). Body language, facial expression and the pheromones are crucial to establishing human relationships, and they are aspects missing in online psychotherapy (Lemma, 2017; Cheshire et al., 2020), which influence the empathy perceived by the patients. However, if therapists have previously seen the patient face-to-face, it can be easier for them to interpret their gestures, expressions, silences or pauses through the screen, which explains why psychotherapists still experience a comparable bond with patients (Leuchtenberg et al., 2022).

Empathy is a reasonably strong predictor of therapy outcome (Elliott et al., 2018), meaning that it can also determine an adequate working alliance between psychotherapist and patient, regardless of the psychotherapeutic approach (Elliott et al., 2011). The importance of empathy in the patient-psychotherapist relationship is widely recognized in literature (Feller and Cottone, 2003; Nascivera et al., 2018).

Sperandeo et al. (2021) studied whether if empathy worked the same way during online sessions. Their results suggest that therapists perceived themselves as being equally capable of providing empathy and support in both settings. Another finding was that the therapists who liked online interventions were mainly those who had previously used this technique. Patients, on the other hand, felt that psychotherapists are more empathetic and more capable of providing support in the online sessions. Due to the pandemic, we must take into account the particular situation in which face-to-face sessions did not offer the same level of comfort as online sessions. At that time, in-person sessions were carried out with masks, plexiglass dividers and strict social distancing. Patients felt much more understood in this setting because they could perceive facial expressions and voice intonations (Sperandeo et al., 2021).

The mixed results regarding the therapeutic alliance may mean that the patient-psychotherapist relationship in online psychotherapy might not yet be optimal (Leuchtenberg et al., 2022). This issue could be addressed with specific training programs for therapists to adapt psychotherapy methods to digital settings, improve the management of technical issues or service evaluation (Leuchtenberg et al., 2022; Lippke et al., 2021; Carrillo et al., 2021). It could help to improve online therapy and make it an effective treatment option to complement face to face therapy, not as a substitute, and also improve the provision of mental health services if we ever need to be prepared to face a situation similar to COVID-19 (Leuchtenberg et al., 2022; Longobardi et al., 2018).

This longitudinal study aims to evaluate the evolution of the therapeutic alliance and face-to-face and online psychological interventions. In this sense, we follow the hypothesis that the face-to-face modality will present a better therapeutic alliance and better patient evolution (both from the patient’s and the therapist’s point of view) between before and after the intervention, compared to the online modality.

A total of 187 subjects participated anonymously and voluntarily in the study, 81 (43.3%) of whom were men and the resting 106 (56.7%) were women. The subjects were aged between 18 and 29 years old, with a mean age of 23.01 (SD 2.82; Table 1).

Initially, a sample of 202 subjects was included, although this was reduced to 187 since 15 of them did not complete the treatment (11 in online treatment and 4 in face-to-face treatment).

The participants came voluntarily and free of charge to the psychological guidance and counseling service, from two universities in Barcelona. At this service, they received counseling and those who were indicated to start a psychological treatment were invited to participate in the study.

The participants responded to the following questionnaires: 1. Sociodemographic data: sociodemographic data such as age, gender, treatment modality (online or face-to-face), and diagnosis were collected ad hoc; 2. Therapeutic alliance—SOFTA-o (System for Observing Family Therapeutic Alliances—observational) for patients (Friedlander et al., 2001); this instrument was created simultaneously in English and Spanish as a transtheoretical tool for research and practice in TA. In this study, the patient version was used. The measure is based on three dimensions: engagement in the process, emotional connection, and safety. It also provides an overall score. The 12 items, both negative and positive, are related to patients’ behaviors, which are grouped within the three dimensions; 3. Clinical Symptomology: The Spanish version (Feixas et al., 2012) of the Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM) is used. It is a questionnaire specifically designed to evaluate the changes that occur in a therapeutic process of patients with various symptoms. It consists of 34 items completed by the patient and it includes four dimensions: (1) Well-being, four items that assess general well-being/malaise; (2) Problems/Symptoms, 12 items that assess anxiety, depression, stress and their somatic manifestations; (3) General functioning, 12 items that assess interpersonal and social relationships and functioning in activities of daily living; and (4) Risk, six items that assess attempts at self-harm, self-harm and aggressive behaviors directed at third parties; (5) HoNOS: The Health of the Nation Outcome Scales (HoNOS), designed by Wing et al. (1998) and translated into Spanish by Uriarte et al. (1999), is an instrument that consists of a set of scales designed to measure the whole range of physical, personal and social problems associated with mental illness and it can be used by Mental Health professionals routinely and in a clinical context. It is composed of 12 items and includes 4 subscales (behavioral problems, deterioration, clinical problems and social problems).

All the subjects filled out the SOFTA-o and the CORE-OM before the intervention began and refilled them once it was concluded. Therapists did the same with the HoNOS.

The interventions lasted between 15 and 20 sessions (45 min per session). The subjects themselves chose whether they wanted to be treated face-to-face or online, because this is a naturalistic observation study and we wanted to know if there was a preference for either of the two modalities. The online interventions were carried out through videoconference.

All subjects were treated by the same therapist (who made the diagnoses guided by their clinical judgment). This therapist is a male psychologist with a PhD in psychology, accredited training in psychotherapy, and over 10 years of experience.

The subjects filled out the questionnaires individually and independently, and they were only assisted by the researcher only when requested.

The study was approved by the Research Ethics Committee of the Vidal i Barraquer Mental Health University Institute.

The statistical analyses were conducted using the SPSS statistical package (version 29.0.0.0(241), SPSS Inc). Firstly, the descriptive results of demographic data, the diagnosis, the TA, CORE-OM and HoNOS were presented. Subsequently, the relation between CORE-OM and HoNOS with intervention modality, gender, age, diagnosis and TA were presented. Afterwards, the mixed model analysis was conducted. To calculate it, an unstructured variance–covariance matrix was computed via the restricted estimation of maximum likelihood. The CORE-OM and HoNOS before and after treatment, treatment modality, and their interactions were considered fixed effects. Finally, TA, gender and age were also included as fixed factors. The random effect was the subjects’ intersection parameter. The degrees of freedom were calculated with the Satterthwaite approximation. The end model was chosen by recalculating the models with and without interaction via maximum likelihood in order to compare the significance of the change on the Akaike information criterion (AIC). The residuals of prediction and of the random factor were inspected via a quartile-quartile plot to assess the suitability of the model.

As seen on Table 1, the percentage of men and women was quite similar, with 81 (43.3%) men and 106 women (56.7%). The mean age was 23.01 (SD 2.83) years old. The 46% (n = 86) chose online modality, while the 54% (n = 101) opted for the face-to-face modality. Therefore, although the distribution is quite similar, there is a certain preference for the face-to-face modality.

The most prevalent diagnosis was anxiety (n = 58.30%), followed by depression (n = 20, 10.7%) and abuse (n = 28.15%).

Regarding TA scores, the SOFTA-o pre-intervention score is 8.58 (SD 3.87) and the post-intervention score is 36.11 (SD 5.34).

As for the CORE-OM scores, it is seen that the mean in the pre-intervention administration was 108 (SD 7.2) and, instead, post-intervention was 75.7 (SD 5.12).

Lastly, the mean HoNOS scores show that, in the pre-intervention administration, the result is 39.65 (SD 3.43), while in the post-intervention administration is 22.68 (SD 2.89).

We conducted, with an IC 95%, a t test for the variables sex and modality; we used the Pearson correlation coefficient for age, TA, CORE-OM and HoNOS and ANOVA for the diagnosis.

Via the Pearson correlation coefficient, as seen in Table 2, in the CORE-OM there is no correlation between the two moments of administration of the instrument (r = 0.057, p = 0.437). Likewise, there is no correlation in the HoNOS between the two moments of administration of the questionnaire (r = −0.011, p = 0.877).

Regarding the CORE-OM scores before treatment, it is shown a significant correlation with HoNOS before treatment (r = 0.195, p = 0.007). And, in reference to the CORE-OM scores post-treatment, there are significant correlations with HoNOS post-treatment (r = 0.447, p < 0.001) and the TA post-treatment (r = −0.471, P = <0.001). The same happens with post-treatment HoNOS scores and post-treatment TA scores (r = −0.882, p < 0.001). These results suggest that patients’ scores on CORE-OM and therapist’s scores on HoNOS are related, meaning there is a relation in the change perceived by therapists and the patients between the pre and post treatment. Even so, therapists perceive a greater improvement after treatment than patients do. Furthermore, it is observed that the CORE-OM and HoNOS scores after treatment (i.e., an improvement in symptomatology) are related to a better TA after treatment.

As seen on Table 3, the t-test for independent samples revealed that there are no significant differences between the CORE-OM before treatment, neither in relation to the modality (t = −0.310, p = 0.378, d = 0.15) nor with the gender (t = −0.182, p = 0.428, d = 0.12). Regarding the CORE-OM scores after treatment, it is shown there is no significant relation with gender (t = −1′605, p = 0.055, d = 0.10) but there is with modality (t = 7.382, p < 0.001, d = 0.66). As for HoNOS, there are no difference between the scores before the intervention in neither modality (t = −0.847, p = 0.199, d = 0.16) nor gender (t = 0.464, p = 0.322, d = 0.04). On the other hand, HoNOS scores after treatment show that there is a significant relation with modality (t = 22.923, p < 0.001, d = 0.91) but not with gender (t = −10.373, p = 0.086, d = 0.07). Therefore, we can conclude that what both patients and therapists report is that face-to-face modality obtains better results than the online modality.

Finally, in relation to the diagnosis, an ANOVA was performed to determine whether there were any differences in the CORE-OM and HoNOS scores by diagnosis, and the results before and after treatment for the two instruments showed non-significant differences. As seen in Table 4, the results for the CORE-OM before and after treatment are F = 2.626, p = 0.170 and F = 3.306, p = 0.278, respectively; and for the HoNOS, F = 1.843, p = 0.194 and F = 2.117, p = 0.446 for before and after the treatment, respectively.

In this case, when comparing the effect size between the variables, it was observed that it was small in all the pairings, since no comparison showed an effect size greater than d = 0.20.

We performed an analysis for CORE-OM and another for HoNOS separately.

Firstly, in the model without interactions, with IC 95%, the pre-post change in the CORE-OM was significant (t₃₇₀ = −38.870, p < 0.001, d = 0.83), as was the treatment modality (t₃₇₀ = −22.483, p = 0.003, d = 0.69). Age, gender, and TA did not reach the level of significance (t₃₇₀ = 0.204, p = 0.87, d = 0.11; t₃₇₀ = −0.098, p = 0.72, d = 0.13; t₃₇₀ = −3.344, p = 0.072, d = 0.14 respectively).

The model with interactions (AIC = 2784.4, with 12 parameters), with IC 95%, was significantly better (X²₄ = 823.42, p < 0.001) than the model without interactions (AIC = 3845.2, with 8 parameters). The interaction between the time of the evaluation and the therapeutic modality was highly significant (t₁₈₃ = −36.974, p < 0.001, d = 0.91). In the web-based treatment, the mean score on CORE-OM decreased 11.24 points (SD 3) while in the face-to-face treatment, it decreased 29.12 (SD 5.2). The rest of the interactions were not significant. In the inspection of the residuals, no gross deviations were found compared to a normal distribution.

In the HoNOS analysis, in the model without interactions, with IC 95%, the pre-post change in the HoNOS was significant (t₃₇₀ = −42.637, p < 0.001, d = 0.88), as was the treatment modality (t₃₇₀ = −31.354, p < 0.001, d = 0.86). Age, gender, and TA did not reach the level of significance (t₃₇₀ = −1.754, p = 0.554, d = 0.6; t₃₇₀ = −1.012, p = 0.487, d = 0.5; t₃₇₀ = −2.876, p = 0.112, d = 0.09, respectively).

The model with interactions (AIC = 2372.6, with 12 parameters), with IC 95%, was significantly better (X²₄ = 745.22, p < 0.001) than the model without interactions (AIC = 4056.1, with 8 parameters). The interaction between the time of the evaluation and the therapeutic modality was highly significant (t₁₈₃ = −43.842, p < 0.001, d = 0.89). In the web-based treatment, the mean score on HoNOS decreased 7.89 points (SD 3.1) while in the face-to-face treatment, it decreased 15.2 (SD 4.2). The rest of the interactions were not significant. In the inspection of the residuals, no gross deviations were found compared to a normal distribution.