Early cancer diagnosis is one of the hardest problems in general practice, because the symptoms are often vague and no single test settles it. AI tools like C the Signs promise to help by analysing the record and flagging cancer risk, and they are now embedded in around 1,400 GP practices. But a risk tool is not the referral standard, and NICE NG12 remains the national framework every GP is accountable to. This is a detailed look at how C the Signs and NG12 each work, what the evidence genuinely shows, where AI risk tools can mislead, and why clinical judgement, including gut feeling, still makes the decision.
In brief: NICE NG12 is the national guideline for recognising and referring suspected cancer in primary care. C the Signs is an AI decision-support tool that analyses the patient record to flag cancer risk and suggest a referral pathway, integrated with EMIS, SystmOne and Vision. The tool supports NG12 and clinical judgement; it does not replace them, and a GP should still refer when NG12 criteria are met, act on gut feeling, and safety-net regardless of a tool's output.
Key takeaways
- NICE NG12 is the national standard for suspected-cancer recognition and referral, organised by site and symptom.
- C the Signs is an AI tool that analyses the record to flag cancer risk and suggest urgent, test-first or non-urgent pathways.
- Reported figures, such as a 98 per cent sensitivity, are largely company or conference-associated and need independent context.
- AI risk tools can help but also risk overdiagnosis, unnecessary referrals and alert fatigue if used uncritically.
- NG12 criteria and clinical judgement, including gut feeling, still decide referral, and safety-netting remains clinical.
What NICE NG12 is
NG12 is the National Institute for Health and Care Excellence guideline on suspected cancer recognition and referral, used across primary care in England and Wales. It sets out, organised by cancer site, symptom and the findings of primary care investigations, when to refer and how urgently, including the two-week-wait pathway and faster thresholds for some presentations. Crucially, NG12 is designed to be used alongside clinical judgement, not mechanically: national bodies are explicit that clinicians should follow the guideline where appropriate while also using their own judgement and gut feeling, use Advice and Guidance when criteria are not met but concern remains, and use primary care investigations to support decisions. NG12 is the standard against which suspected-cancer referral is judged, and it assumes a thinking clinician, not a checklist.
What C the Signs is
C the Signs is an AI-powered clinical decision-support tool, founded by NHS doctors, that analyses a patient's electronic record for combinations of signs, symptoms, risk factors and clinical markers to flag cancer risk and suggest an appropriate pathway, sorting patients into urgent, test-first or non-urgent routes across a wide range of cancer types. It is accessed through a one-click desktop toolbar and integrates with EMIS, SystmOne and Vision, with assessments, safety-netting and tracking coded back into the record, and it can report a practice's cancer performance data such as two-week-wait activity and detection rates. It has been through regulatory approval as a Class 1 medical device with the MHRA. Its stated aim is to help identify at-risk patients earlier while reducing unnecessary urgent suspected-cancer referrals.
What C the Signs does well
The genuine value is real. By analysing the whole record, it can surface combinations of factors a busy clinician might not connect in a ten-minute consultation, prompting consideration of a referral that might otherwise be missed. It embeds safety-netting and tracking into the workflow, coded back to the record, which supports continuity and follow-up. It suggests a diagnostic pathway rather than just a risk score, which is more actionable. And it gives practices and primary care networks visibility of their real-time cancer performance data, which supports quality improvement. For prompting consideration and structuring follow-up, it does a useful job that pure guideline text does not.
The evidence, honestly
The headline numbers are striking and should be read with care. Interim findings presented at a major oncology conference reported a 98 per cent sensitivity for spotting cancer across more than 111,000 risk-assessed patients, and the company has reported a large return on investment, with an independent evaluation conducted in one integrated care board area. These are encouraging, but much of the evidence base is company-associated or presented at conferences rather than drawn from large independent randomised trials, so it warrants the same scrutiny as any vendor-reported outcome. It is also worth remembering the wider history of cancer risk-assessment tools such as QCancer and the RAT, which despite integration into GP systems saw low uptake and raised ongoing questions about whether they improve outcomes without an unacceptable rise in unnecessary investigation. Promising is the right word; proven at scale, independently, is a higher bar.
Where AI cancer tools can mislead
The failure modes deserve as much attention as the benefits. The first is overdiagnosis and over-referral: a tool tuned for high sensitivity will flag many patients who do not have cancer, and acting on every flag risks unnecessary investigation, patient anxiety and pressure on already-stretched diagnostic services. The second is alert fatigue: frequent prompts can dull attention, so that genuine flags are clicked through with the rest. The third is false reassurance: a low-risk output must never override a clinician's concern, because the tool does not see everything and cannot feel the unease that a subtle presentation generates. And the fourth is deskilling: relying on a tool to notice patterns can erode the clinician's own pattern recognition over time. None of these mean the tool is bad; they mean it must be used as support, with judgement intact.
Judgement and gut feeling still decide
This is the heart of it. NG12 and C the Signs both feed into a decision that remains clinical. If a patient meets NG12 referral criteria, refer, regardless of what a risk tool says. If your gut feeling says something is wrong even when criteria are not met and a tool scores low risk, act on it, using Advice and Guidance or investigation to work it through, because gut feeling is a recognised and legitimate part of NG12-aligned practice. And safety-netting, telling the patient what to look out for and when to return, and arranging appropriate follow-up, is a clinical act that no tool performs for you. The tool can prompt and structure; the clinician decides and safety-nets.
C the Signs and NG12: complementary, not competing
The right framing is not tool versus guideline. NG12 is the standard and the reasoning framework; C the Signs is a way of surfacing risk from the record and structuring follow-up within it. Used well, the tool prompts consideration and supports tracking, while NG12 and clinical judgement determine the referral, and the two reinforce each other. Used badly, the tool becomes a substitute for thinking, generating over-referral or false reassurance. The difference, again, is whether the clinician stays in charge.
Where iatroX fits
C the Signs answers "which patients might be at risk"; the harder questions are "why, and what should I do". iatroX sits in that reasoning layer: when a presentation raises a suspected-cancer question, Ask iatroX gives guideline-grounded answers from NICE, CKS, SIGN and the SmPC with the source attached, so you can check the NG12 threshold, work through the differential, decide what to safety-net, and learn from the case for practice and exams. It is not a risk-scoring tool, and it does not replace NG12 or your judgement; it helps you reason to the right decision and retain the learning. Try it at Ask iatroX, and for the uncertain cases where referral criteria are not met, see how to ask for Advice and Guidance as a GP. For how these tools sit together in practice, see the hidden GP software stack.
Frequently asked questions
What is the difference between C the Signs and NICE NG12? NG12 is the national guideline defining when and how urgently to refer suspected cancer. C the Signs is an AI tool that analyses the record to flag cancer risk and suggest a pathway. NG12 and clinical judgement decide the referral; the tool supports that decision.
Is C the Signs accurate? Reported figures, such as a 98 per cent sensitivity across a large patient sample, are encouraging but largely company or conference-associated, so they warrant independent scrutiny. Treat it as promising decision support, not a proven substitute for clinical judgement, and remember the mixed history of cancer risk tools.
Should I refer if C the Signs says low risk but I am worried? Yes. A low-risk output must never override clinical concern. If your gut feeling says something is wrong, or NG12 criteria are met, act on it using referral, investigation or Advice and Guidance. Gut feeling is a legitimate part of NG12-aligned practice.
Can AI cancer tools cause harm? They can, if used uncritically. Risks include over-referral and unnecessary investigation from high-sensitivity flagging, alert fatigue, false reassurance from low-risk outputs, and deskilling. Used as support with judgement intact, they help; used as a substitute for thinking, they can harm.
Does C the Signs replace clinical judgement? No. It surfaces risk and structures follow-up, but the referral decision, the response to gut feeling, and safety-netting remain clinical acts. NG12 assumes a thinking clinician, and the tool is there to support that clinician, not replace them.