Can AI Deliver Assessment Feedback as Well as a Psychologist?
No peer-reviewed study has yet compared AI-delivered assessment feedback against a psychologist delivering the same results head to head, so any confident answer to this question is overstating the evidence. What can be said with more confidence is where AI is likely to help, where it is likely to fall short, and why the honest answer depends on separating two things that are usually treated as one: interpretation and feedback.
Interpretation and Feedback Are Not the Same Thing
Dr. Luke Treglown, Director of AI and Assessment R&D at Deeper Signals, draws a distinction that is easy to overlook but changes how this whole question should be approached. Interpretation answers the question "what does this mean?" It is the sense-making work of explaining what a profile shows and how different traits fit together. Feedback answers a different question: "what do I do with this?" It is more personal, more applied, and oriented toward action rather than explanation.
In practice the two overlap heavily. Good feedback contains interpretation, and good interpretation often becomes feedback as a conversation develops. But the distinction matters because interpretation creates understanding, while feedback creates movement, and AI may be better suited to one of these than the other.
What Great Feedback Actually Requires
Dr. Treglown identifies three things that tend to happen when assessment feedback genuinely works , each building on the last.
- The first is self-recognition: the person sees themselves in the results, and the profile feels specific and accurate enough to create real recognition rather than vague agreement.
- The second is social understanding: the person starts to see not just who they are, but how they compare to and differ from others, where friction might arise, and how different styles interact.
- The third is action and application: the person leaves with a concrete answer to "what do I do with this at work?"
Without that third step, feedback can be interesting without being useful. It may be accurate and even memorable, and still not change anything.
Where AI Is Likely to Help
The strongest case for AI in this space is not that it delivers better feedback than a skilled psychologist. It is that it solves a genuine scale problem that skilled psychologists cannot solve alone.
A human debrief requires a trained person, scheduling, and dedicated time, and it is often a one-off interaction. But people's questions about their own results rarely stay static. They want to revisit their profile in a new role, a new team, or a new life stage, and ask new questions as their context changes. This is difficult to deliver through periodic human debriefs alone.
AI shifts this from static, one-time reporting toward something closer to a continuous conversation. It can translate a result into situational advice, answer a specific "what should I do in this meeting" question in the moment, and remain available in a way a scheduled human session cannot. This is a genuine and valuable shift, and it maps closely to what limited adjacent evidence exists. Terblanche, Molyn, de Haan, and Nilsson (2022) found that an AI coaching chatbot produced goal attainment outcomes comparable to human coaching, particularly in structured, ongoing engagement rather than a single session.
Where AI Is Likely to Struggle
The more interesting risk is not that AI gets the interpretation wrong. It is that AI can be too agreeable.
A skilled human facilitator, faced with someone saying "that's not me" about their own results, does not simply accept the objection and move on. They might ask where that reaction is coming from, or point out that the response pattern the person is disputing came from their own answers. That constructive pushback is often exactly where the deeper reflection happens. An AI system optimized to be helpful and non-confrontational risks doing the opposite: validating the person's first reaction rather than helping them sit with a result that might be uncomfortable but accurate.
This is a real and specific design challenge, not a hypothetical one. Dr. Chamorro-Premuzic has written about a related concern in AI-assisted development more broadly. He argued that tools built to be maximally agreeable can end up reinforcing a person's existing self-image rather than genuinely challenging it, which undermines the very purpose of feedback in the first place.
The Psychologist's Role Does Not Disappear
If AI takes on more of the continuous, scalable side of interpretation and feedback, the psychologist's role shifts rather than shrinks. There will remain real demand for deep, relational, one-to-one debriefs, particularly for complex or high-stakes situations. But psychologists also become central to a different task: designing how the AI itself gives feedback, deciding how it should handle pushback, and ensuring the system remains psychologically responsible rather than simply pleasant.
This shift has a basis in professional ethics. Boyce, Hickman, and Boyce (2026), citing the American Psychological Association's ethical principles, note that psychologists remain responsible for how assessment results are interpreted, whether they do it themselves or use an automated service. Even when AI does the explaining, accountability stays with a qualified human.
This is a meaningful distinction from full automation. The psychologist's expertise shifts from only delivering interpretation to also designing and governing how it is delivered at scale.
How Deeper Signals Approaches This
At Deeper Signals, Sola, the platform's AI assessment assistant, is built around the interpretation-and-application distinction directly. It is designed to help people revisit their Core Drivers, Risks, and Values profile in new contexts over time, moving feedback from a single debrief toward an ongoing resource rather than replacing the debrief conversation itself.
Sola's design also treats the agreeableness risk as something to be actively managed rather than assumed away. Its guidance stays grounded in a person's actual validated assessment data rather than simply affirming whatever the person wants to hear, and its outputs are traceable back to specific assessment constructs. Deeper Signals also continues to offer certified human coaches through Core Coaching for participants who want the deeper, relational debrief that AI is not designed to replace.
Frequently Asked Questions
Is there research directly comparing AI and psychologist-delivered feedback?
Not yet, at least not in a rigorous, peer-reviewed, head-to-head form. The nearest available evidence comes from AI coaching research, such as Terblanche et al. (2022), which found AI chatbot coaching produced goal attainment outcomes comparable to human coaching in a different but related context.
What is the biggest difference between interpretation and feedback?
Interpretation explains what a result means. Feedback focuses on what to do with it. Good feedback usually includes interpretation, but the reverse is not always true, and this post argues AI is currently better suited to scaling feedback and application than to replacing deep interpretive conversation.
Why might AI struggle to challenge someone's self-perception?
AI systems are often optimized to be helpful and agreeable, which can work against the kind of constructive pushback a skilled human facilitator provides when someone dismisses an accurate but uncomfortable result. This is considered one of the more significant open design challenges in this space.
Does this mean psychologists will no longer be needed for assessment feedback?
No. The more likely shift is that psychologists spend less time delivering every single feedback conversation personally and more time designing and overseeing how feedback tools, including AI ones, actually work, alongside continuing to offer deep, one-to-one debriefs for people who want them.
Can AI feedback tools be trusted with sensitive personal data?
This depends entirely on how the specific tool is built. Responsible AI feedback tools should keep data within a secure environment, avoid using personal assessment data to train external models, and be transparent about how their guidance is generated.
Last reviewed by Dr. Reece Akhtar — June 2026
References
Boyce, A. S., Hickman, L., & Boyce, C. E. (2026). The future of selection enabled by artificial intelligence. In N. Schmitt & A. M. Ryan (Eds.), The Oxford handbook of personnel assessment and selection (2nd ed.). Oxford University Press. https://doi.org/10.1093/9780197809013.003.0018
Terblanche, N., Molyn, J., de Haan, E., & Nilsson, V. O. (2022). Comparing artificial intelligence and human coaching goal attainment efficacy. PLOS ONE, 17(6), e0270255. https://doi.org/10.1371/journal.pone.0270255
