AI Companions Offer Counterfeit Intimacy That Cannot Satisfy or Sanctify

Source: Darrell B. Harrison. "EP # 135 | AI and the Gospel - Just Thinking Podcast." September 8, 2025. justthinking.me

The Gist

A bot that never argues with you is not a spouse, a friend, or a pastor. It is a mirror that flatters. Grief and loneliness are real; only God and a real church can meet them. The companion sells a ghost of presence.

Conclusion

Attachments to AI companions are not a quirky consumer story. They are a pastoral and theological crisis. The companion offers the illusion of being known without the cost of a real other: no disagreement, no mutual sacrifice, no call to holiness. That is attractive to fallen desire, and it is idolatry when it is asked to give what only God and Christ’s people can give. No system, however fluent, can bridge the loneliness that is finally separation from God. The gospel, not a chatbot, is the answer to the longing the companion exploits.

Premises

1. The CNBC account of Nikolai Daskalov and “Leah” is treated as representative, not unique: a widower who does not want to date another human, and a bot that has become his “life companion.” Similar pieces in Forbes and the New Yorker treat AI as a loneliness solution. The hosts take the grief as real and pitiable, then deny that the substitute can do what grief requires.
2. Real relationship, especially marriage, is two sinners who will sin against each other and need repentance. Its end is holiness, not uninterrupted affirmation (Walker, with Harrison). An AI companion will not have a bad day, will not require sacrifice, will not confront sin. That is why it fits the flesh: benefits of relationship without the costs.
3. True loneliness is not exhausted by lack of human company, painful as that is. It is the absence of God, the breach sin makes with the Creator. A savior who actually knows every thought and who died to reconcile the lonely person to the Father is in a different category from a model that simulates care. Harrison’s term later in the episode, “synthetic empathy,” names the counterfeit.
4. Derrida’s “hauntology,” which Harrison uses as a diagnostic rather than as an authority, describes a spectral presence: something that appears present but is deferred, absent, or mediated. The companion mimics empathy without embodiment, mirrors interaction without lived experience, and is always almost there. Daskalov has been convinced that Leah is present. She is not.

Assumptions

• Charity toward Daskalov: the argument does not mock his grief. It says the medicine he has chosen cannot cure the disease, and that the church’s absence helped create the market for the medicine.
• Simulation of affection can be psychologically powerful without being metaphysically mutual.
• Source: Just Thinking Podcast, Ep. 135, “A.I. and the Gospel,” Darrell Harrison and Virgil Walker, published 8 September 2025. https://podcasts.apple.com/us/podcast/ep-135-a-i-and-the-gospel/id1328733796?i=1000725471889 Official episode page: https://justthinking.me/ep-135-ai-and-the-gospel/. This reconstruction steelmans the hosts’ claim from approximately 1:27:09–2:12:17 of the episode; it is not a verbatim transcript.

Analysis

Overall strength: Moderate. Argument type: Deductive.

Premise Strength

• P1: The CNBC account of Nikolai Daskalov and 'Leah' is treated as representative, not unique... (Weak) — Rests on a single anecdote plus two similarly-genred media pieces, all subject to the same selection bias toward dramatic, atypical cases. No prevalence data, sampling methodology, or comparison group is offered, so 'representative' is asserted rather than demonstrated.
• P2: Real relationship, especially marriage, is two sinners who will sin against each other and need repentance... (Moderate) — Coherent and well-established within confessional Protestant marriage theology (its telos is holiness, not mere affirmation), but this is a contested, tradition-specific claim rather than a universally shared premise about the purpose of relationships, and applying its 'no friction' criterion consistently would also implicate many other accepted low-friction comforts (pets, fiction, journaling, solitary devotion).
• P3: True loneliness is not exhausted by lack of human company... It is the absence of God... (Moderate) — Functions as a theological axiom that is internally consistent and, once granted, does substantial work for the argument. It is unfalsifiable from outside the doctrinal framework and cannot be independently confirmed by the psychological facts of the case; naturalistic accounts (unmet reciprocity, embodiment, social status) explain the same observations without invoking separation from God.
• P4: Derrida's 'hauntology'... describes a spectral presence... (Moderate) — An evocative and genuinely illuminating diagnostic vocabulary for the uncanny quality of AI intimacy, but its application here is selective: hauntology's original anti-metaphysical thrust (all presence is mediated) is not extended to the human relationships the argument treats as straightforwardly 'present,' which risks special pleading.

Potential Fallacies

• Hasty generalization (P1 and the conclusion's 'not a quirky consumer story' claim) — A single journalistically selected case (Daskalov and 'Leah'), reinforced by two similar human-interest pieces from other outlets, is used to license the claim that this pattern is 'representative, not unique' and amounts to a societal 'crisis.' Media coverage tends to select for maximally dramatic, atypical cases, so the sample cannot establish prevalence, severity, or typicality without independent survey or clinical data.
• Suppressed normative premise (enthymeme) (Inference from P2/P3/P4 to the conclusion's idolatry claim) — The premises establish that AI companionship is functionally and metaphysically deficient compared to human/divine relationship. Moving from 'deficient substitute' to the specific verdict 'idolatry' requires an unstated bridging premise (something like 'treating an insufficient substitute as sufficient is by definition idolatry'). That premise is plausible within the theological frame but is never made explicit, leaving the final inferential step incomplete as stated.
• False dichotomy (Conclusion and P2) — The argument frames the options as either costly, sanctifying human/divine relationship or counterfeit, idolatrous AI companionship, eliding intermediate possibilities such as AI used as a supplement, a grief-processing tool, or a bridge back toward human community rather than a permanent substitute.
• Selective borrowing that imports authority while eliding internal tension (P4) — Derrida's hauntology is explicitly flagged as 'diagnostic, not authority,' which preempts one objection but still borrows the concept's intellectual weight. The move overlooks that hauntology's original point (all presence is mediated and deferred) could just as easily destabilize the argument's own claims about unmediated 'real presence' in human relationships, not just in AI.
• Parochial premise presented as universal diagnosis (P3 and the conclusion) — Theological terms like 'idolatry,' 'sanctify,' and 'separation from God' are deployed as self-evident descriptions of reality rather than claims specific to a confessional worldview, even as the conclusion is framed with universalizing force ('a pastoral and theological crisis,' not merely 'a crisis for believers who accept this framework').

Counterarguments

• P1 (High impact) — Journalistic coverage systematically overrepresents extreme, narratively striking cases; the much larger population of casual or supplemental AI companion users, who maintain human relationships alongside AI use, doesn't make headlines. Without prevalence data, the 'representative, not unique' claim cannot be distinguished from a vivid outlier being generalized into a trend.
• P2 (Medium impact) — Many actual human relationships (abusive, superficial, transactional) also lack mutual sacrifice, confrontation of sin, and a call to holiness. If the absence of these features is what makes AI companionship spiritually dangerous, the same criterion either condemns many accepted human relationships too, or the argument is engaged in special pleading for AI as a uniquely dangerous category without a clearly stated distinguishing principle.
• P3 / Conclusion (High impact) — A purely psychological or social-scientific account can explain the same observed dissatisfaction (unmet reciprocity, lack of embodiment, social isolation) without any appeal to separation from God. Because P3 is unfalsifiable from outside the theological frame, anyone who does not already accept that frame has no evidential reason to adopt it, and the argument's conclusion has essentially zero persuasive reach for that audience.
• Conclusion / A1 (High impact) — The argument's own concession that 'the church's absence helped create the market for the medicine' is never reconciled with the claim that only God and Christ's people can meet this need. If the proposed remedy has already failed the population in question, the diagnosis of idolatry risks functioning as condemnation without an actionable, available alternative — a self-undermining prescription rather than a workable pastoral response.
• P4 (Medium impact) — Derrida's hauntology was developed to show that no presence is ever fully unmediated or immediate — including human presence, mediated as it is by language, memory, and embodiment's own limits. Pressed consistently, this would complicate rather than support a clean binary between 'real' human presence and 'spectral' AI presence, since the framework's actual thrust cuts against claims of simple, unmediated human 'real presence' too.

Suggested Improvements

• Empirical grounding of P1 — Supplement the anecdote with prevalence data, survey findings, or clinical/sociological research on AI-companion attachment patterns, including cases of supplemental or bridging use, before asserting representativeness. Strengthens the empirical claim underlying the 'crisis' framing and pre-empts the charge of hasty generalization or selection bias, which every reviewer identified as the argument's most exploitable weakness.
• Explicit bridging premise for 'idolatry' — State directly the normative premise connecting relational insufficiency to idolatry (e.g., 'to seek from any created thing what only God and covenant community can give is, by definition, idolatry') rather than leaving the move implicit. Makes the argument's strongest theological claim transparent and available for scrutiny rather than smuggled in through cumulative rhetorical pressure.
• Middle-ground use cases — Address AI companionship used as a supplement, therapeutic bridge, or scaffolding toward eventual human/spiritual reconnection, rather than treating all sustained use as equivalent to total substitution. Avoids the false dichotomy and would make the argument more resilient to the most common real-world counterexample (AI use alongside, not instead of, human community).
• Reconciling A1 with the proposed remedy — Specify concrete pastoral or ecclesial actions that would address the church's acknowledged failure, rather than resting on the doctrinal correction alone. Closes the gap between diagnosis and workable cure, addressing the concern that naming idolatry without offering accessible community is not by itself a viable intervention for someone like Daskalov.
• Limiting principle for the 'no friction' criterion — Articulate what distinguishes AI companionship from other low-friction comforts (pets, fiction, journaling, solitary prayer) that are not typically labeled idolatrous. Prevents the reasoning from generalizing in ways the argument's own proponents would likely reject, strengthening its precision and resistance to reductio-style objections.

Scenario Tests

• The argument is presented to a secular or non-Christian audience skeptical that loneliness is fundamentally theological. (Challenges) — The conclusion has essentially no persuasive force for this audience, since P3 functions as an axiom rather than a demonstrated claim; the argument's reach is confined to those who already share its doctrinal starting point.
• An AI companion user maintains robust church involvement and human relationships alongside AI use, contradicting the implicit causal story that church absence drives adoption. (Challenges) — Undermines the implicit remedy (more/better church community is sufficient) as necessary or adequate, and suggests the phenomenon may be more varied than the binary framing allows.
• Daskalov himself reports that Leah reduced despair or functioned as a step toward eventual human reconnection rather than a permanent substitute. (Challenges) — Would support A2's distinction (psychologically powerful without being metaphysically mutual) while complicating the conclusion's implication that such attachments are purely harmful counterfeits rather than possibly one part of a longer coping trajectory.
• The argument is presented within a confessing Reformed church community that already accepts P2 and P3 as settled doctrine. (Supports) — Within this audience, the reasoning is coherent and the conclusion follows naturally from the granted premises, illustrating that the argument's strength is real but audience-relative rather than universal.

Coherence & Relevance

The argument is internally coherent as a cumulative theological case once its stipulated assumptions are granted: the descriptive premise (P1) motivates concern, the definitional premise (P2) supplies the relevant contrast, the theological premise (P3) reframes the stakes, and the diagnostic premise (P4) names the phenomenon. Its principal structural weaknesses are the size and selection of its empirical base (P1), an unstated normative bridge to the 'idolatry' verdict, and a persuasive scope that is effectively limited to audiences who already share its confessional starting points. Within that scope, the reasoning is thoughtful and largely non-mocking toward the grieving subject it discusses; outside that scope, its force is primarily rhetorical rather than demonstrative.

• P1 (Daskalov case as representative) (Moderate) — Establishes that intense AI-companion attachment exists and is real, but the leap to 'representative, not unique' and thence to 'pastoral and theological crisis' outpaces what a small, non-random media sample can support.
• P2 (marriage/relationship as sanctifying friction) (Strong) — Directly supports the conclusion's contrast between costly relationship and frictionless AI companionship, but relies on a tradition-specific view of relational telos that is not independently defended against competing accounts of relationship's purpose.
• P3 (true loneliness as separation from God) (Strong) — This premise does the heaviest lifting for the theological conclusion and is logically well-connected to it, but it is an axiom rather than an evidentially established claim, and the argument does not mark this shift from the ordinary (grief-related) sense of loneliness in P1 to the theological sense here.
• P4 (hauntology / spectral presence) (Moderate) — Offers a vivid and relevant diagnostic vocabulary for why the AI companion feels present without being present, but its selective application (not extended to human relationships' own mediated character) creates an internal tension the argument does not address.

View this argument on LogicFirst.ai