The Rise of AI in Dentistry: How ProAligner Uses Algorithms to Plan Your Smile

Aligner treatment looks simple from the outside — a series of clear trays, each slightly different from the last. What is not visible is the planning that produces that series, and over the past decade that planning has shifted almost entirely from a technician's judgement to software.

The word "AI" gets used loosely in dental marketing, sometimes to describe genuine machine learning trained on large clinical datasets and sometimes to describe fairly ordinary geometric software. It is worth understanding which is which, because the distinction affects what you can reasonably expect from a digital treatment plan.

This article explains what the algorithms genuinely do, what they are good at, where they fall short, and why a registered clinician remains responsible for every stage.

What does the software actually do?

It takes a three-dimensional scan of your teeth, identifies each tooth as a separate object, estimates the position of the roots beneath the gum from radiographic and anatomical data, and then calculates a sequence of small movements that takes each tooth from where it is to where the clinician has prescribed it should be. Each stage in that sequence becomes one aligner. The software also flags movements it judges to be unpredictable, so the clinician can add attachments or plan differently.

Orthodontic planning used to begin with alginate impressions and plaster models. A technician would physically saw the model into individual teeth, reposition them in wax, and produce a target arrangement by hand. It worked, but it was slow, the accuracy depended entirely on the technician, and there was no straightforward way to plan the intermediate stages.

Intraoral scanning replaced the impression tray, capturing thousands of surface points and producing a digital model without material setting in your mouth. Our comparison of digital scans and putty impressions covers the practical differences from a patient's point of view. Once the model exists as data, everything downstream becomes computational: segmentation, movement planning, staging, and manufacture of the trays themselves.

The first task is separating a single continuous scan of an arch into discrete teeth. Machine learning models trained on large numbers of annotated scans now do this reliably, identifying tooth boundaries at the gum margin and between adjacent contacts. The software then estimates root morphology, either from population averages for each tooth type or, where a CBCT scan is available, from your actual anatomy.

Given a starting position and a target position, the software calculates a path. This is less trivial than it sounds, because teeth cannot pass through each other and space has to be created before certain movements can happen. The algorithm sequences movements so that, for example, space is opened before a rotated tooth is derotated into it, and it limits each stage to a movement small enough to be biologically achievable — typically a fraction of a millimetre.

Not all movements respond equally well to aligners. Bodily movement of roots, extrusion of a tooth downwards, and rotation of round-rooted teeth such as canines and premolars are known to be less predictable. Planning software applies weightings based on accumulated clinical data, over-correcting some movements and prompting the clinician to add composite attachments that give the aligner something to push against.

The output includes an animated preview of the projected sequence. This is genuinely useful for communication, but it should be understood for what it is: a rendering of the plan, not a photograph of the future. Our article on what smile simulations can and cannot show explains where previews are reliable.

Teeth move because sustained light pressure triggers bone remodelling — resorption on the pressure side, deposition on the tension side. The rate is biological and roughly a millimetre a month; software cannot accelerate it. What software can do is optimise the sequence so movements happen in an efficient order and forces stay within the range that produces movement rather than stalling it.

Where genuine machine learning contributes most is in prediction. Systems trained on large numbers of completed cases can identify which planned movements historically fell short of their target and by how much, allowing the plan to compensate in advance. This is why modern aligner systems generally track better than early ones did — not because the plastic changed dramatically, but because the planning learned from outcomes.

An algorithm optimises a path towards a target. It does not decide whether that target is appropriate for you. Deciding what the teeth should look like, whether your gums can withstand the planned movement, whether extractions are needed, whether the bite will function afterwards, and whether aligners are the right appliance at all — these are clinical judgements made by a GDC-registered dentist after examining you.

Software also cannot diagnose. It does not see decay, it does not measure periodontal pockets, and it does not read radiographs for pathology. Starting orthodontic treatment on a mouth with active gum disease or untreated decay risks making both worse, which is why an in-person examination with radiographs must come before any digital plan is accepted.

• Root positions are frequently estimated rather than measured

• Bone density and periodontal condition are not captured by a surface scan

• Individual biological response varies and cannot be predicted from geometry

• Soft tissue behaviour — lips, gums, tongue posture — is poorly modelled

• The plan assumes full compliance with the prescribed wear schedule

Where reality diverges from the plan, the clinician detects it at review appointments and corrects course. That is precisely why in-person reviews matter and why treatment without them is a different proposition.

Before any digital planning, an assessment should establish:

• Whether your teeth and gums are healthy enough for orthodontic force

• What is actually causing the concern — crowded teeth, gaps between teeth, or a bite discrepancy

• Whether aligners can achieve what you want, or whether fixed appliances are more appropriate

• What the risks are in your particular case

• What retention will involve afterwards

Our ProAligner page explains what that assessment involves and how planning proceeds from it.

Wear time. Aligners work only while they are in. Most protocols require 20 to 22 hours a day, and shortfalls accumulate — a tray that has not achieved its movement means the next one does not seat properly, and tracking is lost.

Attend reviews. The plan is a projection; reviews are where divergence is caught. Missing them is how small problems become refinement stages.

Keep the mouth clean. Aligners hold saliva and anything else against the teeth. Brush before reinserting, clean between teeth daily, and keep up hygiene appointments.

Plan for retention. Whatever the software achieved, teeth drift back without retention. A fixed retainer or nightly removable retainer is a permanent part of the arrangement.

• Planning software segments a digital scan and sequences achievable movements into aligner stages

• Machine learning contributes mainly through prediction of how movements historically perform

• Algorithms cannot diagnose, and root position is often estimated rather than measured

• The biological speed of tooth movement is unchanged by software

• A registered clinician prescribes the target, assesses suitability and supervises progress

• The animated preview is a plan, not a prediction of your finished result

• Wear time and attendance at reviews determine how closely reality follows the plan

The NHS overview of orthodontics provides neutral background on what orthodontic treatment involves.

Digital planning has made aligner treatment more predictable and considerably easier to explain to patients. It has not changed the biology, and it has not removed the need for a clinician to examine you, decide whether treatment is appropriate, and take responsibility for the outcome.

If you are considering aligners, the question to ask is not which software a provider uses, but who is diagnosing your case, how often you will be seen, and what happens if the teeth do not follow the plan.

To discuss whether aligner treatment suits your case, arrange an appointment at 22 Wimpole St, London W1G 8GQ, or call 020 7183 0692.

Dental symptoms and treatment options should always be assessed individually during a clinical examination.

This article is for general educational information only and is not professional dental advice, diagnosis, or a treatment recommendation. Information here is general and cannot replace an in-person assessment by a qualified, GDC-registered dental professional. Symptoms, suitability, fees, timelines, and outcomes vary according to individual clinical circumstances. If you have pain, swelling, or other concerning symptoms, seek prompt professional dental care. Always request a written treatment plan and cost estimate before proceeding with treatment.

Written Date: 3 August 2026 Next Review Date: 3 August 2027