CQC Mock Inspection Template: Run One and Summarise Findings With AI

A care home can run a CQC mock inspection by checking each of the five key questions (safe, effective, caring, responsive, well-led) against a short template, capturing evidence for every finding, then using AI to cluster the notes into themes and draft an action plan. A manager still checks and owns every finding.

The value of a mock is not the score. It is the list of things you would not have noticed: the care plan review that slipped, the induction record with no date, the policy nobody has read since it was written. AI helps at the end, turning pages of observations into a ranked, owner-and-date action plan in minutes. It cannot do the walking round, the talking to people or the reading of records, so the template has to make those steps easy to do properly.

Follow me on Instagram@sagnikteaches

Check which framework wording applies on the day

CQC's assessment framework, as its pages stood in late October 2026, has five key questions and, under each, quality statements written as "we" statements: 34 in all (8 safe, 6 effective, 5 caring, 7 responsive, 8 well-led). Evidence is grouped into six categories: people's experience of health and care services, feedback from staff and leaders, feedback from partners, observation, processes and outcomes. A care home mock should touch all six, which is why the day below mixes walking round, reading records and talking to people.

Connect on LinkedInSagnik Bhattacharya

That structure is being replaced. In 2026 CQC consulted on four draft sector-specific frameworks, one of them for adult social care, and the consultation closed on 12 June 2026. In the drafts the five key questions stay, ratings are made at key-question level without scoring, and the quality statements give way to key lines of enquiry (between three and seven per key question), alongside rating characteristics and "I statements" describing what people expect. I found no start date on CQC's pages, so open the framework page for adult social care before each mock and copy the wording in force. The template below is built on the five key questions, so it survives the change: only the statement labels on each row need to move to the new lines of enquiry.

Subscribe on YouTube@sagnikteaches

The mock inspection template, by key question

Use one page per key question. Each page has the same five columns: what you check, how you check it, who you ask, evidence found and a rating of strong, partial or weak. Here are six filled-in rows, one or two per key question (fictional home, illustrative findings).

Label every row with the quality statement it evidences, using CQC's exact names: the rows above map to medicines optimisation, involving people to manage risks, consent to care and treatment, kindness, compassion and dignity, listening to and involving people, and governance, management and sustainability. Staff training sits under safe and effective staffing in the current framework, not under effective as it did in the older key lines of enquiry, which is a common mislabelling in templates carried over from years ago. Build 6 to 10 rows per key question, drawn from the things CQC's guidance for your service type says it looks at and from your own risk areas. Sampling matters: check a number of records, not one, and state the sample size in the finding. A statement such as "2 of 8" is far more useful to AI and to a manager than "mostly fine".

Plan the day: who walks, who reads, who talks

A single-home mock takes about a day for someone experienced. Split the work into three strands, run in parallel if you have more than one reviewer.

1. Observation (2 hours). Walk every floor at different times. Watch a mealtime, a medicines round and a handover. Note what you see, not what you infer.
2. Records (3 hours). Sample care plans, risk assessments, medicines records, staff files, incident logs, complaints and audits.
3. Conversations (2 hours). Talk to four residents or relatives, four staff across roles and one visiting professional if possible. Ask open questions such as "What would you change here?" and "Who do you go to if something worries you?"

Keep your voice recorder and phone away from people's private spaces. If you take notes on a device, type themes, not names. Every finding gets a code, for example S-03 for the third finding under Safe, so evidence can be traced without naming people.

Capturing evidence in a way an AI summary can use

Notes written as sentences are hard to compare. Notes written in a fixed shape are easy to cluster. Use this shape for every finding:

Code: S-03
Key question: Safe
Area: Medicines
What I checked: 8 MAR charts, two units, last 14 days
What I found: 2 of 8 had an unexplained gap on Saturday night shift
Evidence reference: MAR chart sample sheet, row 4 and 6 (no names)
Strength of evidence: records only, not yet discussed with staff
Risk if repeated: missed doses
Suggested owner: Deputy manager

The "strength of evidence" line stops you treating a single record check as a verified fact. The reference points back to a document in your own system rather than carrying personal details into a tool.

Giving findings to AI without giving it people

De-identify before you paste. Remove names, room numbers, dates of birth and anything that could single out a person; refer to "Resident A" if you must, and better still, record themes and counts instead. If your home has a tool approved for personal data with a data agreement in place, your policy decides what may go in; otherwise assume anything pasted into a general AI tool has left your control. The staff policy for this is covered in AI in care homes: a staff policy covering data protection, consent and CQC.

A prompt for the summary step:

You are helping a registered manager after a mock inspection. Below are findings in a fixed format, with no personal data. Group them by key question (safe, effective, caring, responsive, well-led). For each group give: the main themes in plain English; the three findings most likely to put people at risk; any finding that appears in more than one key question. Do not invent findings and do not rate the home. If two findings look the same, say so. Mark anything where the evidence is "records only".
Findings: [paste]

Sample output, illustrative:

Safe: Two themes. Medicines recording gaps at weekends (S-03, S-07). Falls follow-up inconsistent (S-04).
Highest-risk: S-03 (missed doses possible), S-04 (one fall without a reassessment), S-09 (door alarm test not logged for three weeks).
Cross-cutting: S-03 and W-02 both point to weekend supervision.
Evidence note: S-03 and S-09 rest on records only.

What you would fix: the AI links S-03 to W-02 (weekend supervision) as a pattern. That is a sensible hypothesis, but you would verify it by asking who was in charge on the two gap weekends before writing it into the plan.

Turning themes into an action plan with owners and dates

An action plan is a short table, not a document. Each action has a finding code, a plain action, an owner by role, a date and a check. Ask the AI for a first draft and then edit it yourself.

Draft an action plan from these themes. Table columns: finding code, action in one sentence starting with a verb, owner (by role), due date (within 7, 30 or 90 days), how we will know it is done. Put anything that could put a resident at risk in the 7-day group. Do not add actions that are not supported by a finding.

A realistic result, edited:

Notice the "done when" column. A plan that says "improve weekend medicines checks" cannot be shown to be finished. A plan that says "next four weekends show no gaps" can.

A worked example: one mock in a 32-bed home

Say a 32-bed home with a registered manager, a deputy and a clinical lead runs its first mock using this template. A reviewer from a sister home spends one day on site. She captures 41 findings: 9 under Safe, 6 Effective, 7 Caring, 8 Responsive, 11 Well-led.

• Raw notes to summary: typing the 41 coded findings takes the reviewer about two hours. The AI summary and draft plan take ten minutes to produce and about forty to edit.
• What the summary found: five themes, two of them crossing key questions (weekend supervision; audit follow-through).
• What the manager changed: merged two duplicate actions, moved one action from 30 to 7 days because it concerned a fire door, and removed an AI-suggested action to "introduce a new falls policy" that the home already had.
• Outcome after 30 days: 11 of 17 actions closed with evidence; the rest were re-dated with a reason.

The numbers are an illustration. What is real is the pattern: raw capture is the slow part, summary is the fast part, and the manager's edit is the part that makes it safe.

A mealtime observation, written so it can be used

Observation is the strand most mock inspections under-record, because it feels subjective. Record times and actions instead of impressions. Illustrative notes from a 40-minute lunch service:

12:10 Dining room, 18 residents seated, 4 staff present.
12:14 Two residents at table 3 waiting for meals; others served.
12:19 Staff member sits beside a resident and cuts food, asks "Would you like it smaller?"
12:25 Resident at table 3 still waiting; staff member at the hatch.
12:29 One resident asks for a drink; carer says "in a minute"; drink arrives 12:36.
12:41 Pureed meal served with each item separate and named by the carer.
Overall: friendly interaction; delay at table 3; no choice offered at the point of service (menu on wall, small print).

Those lines become findings with a code and a sample size of one meal, which is honest. If three different mealtimes show a delay at the same table, you have a pattern. The AI summary can only cluster what you captured, so the quality of the observation notes decides the quality of the plan.

Questions to ask residents, relatives and staff

Open questions draw out more than "Are you happy here?". A short bank to adapt:

• Residents: "What is a good day here?" "If you rang the call bell now, how long would you expect to wait?" "Is there anything you would like to do that you cannot?"
• Relatives: "How do you hear about changes in your relative's care?" "When did you last raise a worry, and what happened?"
• Care staff: "What would you do if you saw a colleague being rough with a resident?" "When did you last have supervision, and was it useful?" "What gets in the way of spending time with residents?"
• Visiting professionals: "How easy is it to get the information you need from the team?"

Write answers as short themes, not names: "Staff said agency cover on nights means fewer who know residents." Tell people before you start that nothing is attributed to them. People answer more honestly when the notes will not carry their name.

Rewriting a weak finding before it goes in the plan

Findings written in a hurry are vague. A before and after shows the standard to aim for.

Weak: "Care plans need work."

Better: "5 of 10 care plans sampled had a review date more than 30 days overdue; 3 of those 5 related to residents with a recent fall."

The second version has a sample, a measure and a link to risk, so an action follows naturally: clear the five overdue reviews in seven days, starting with the three after falls. When you ask AI to tidy a batch of findings, tell it to keep the numbers exactly and not to soften them. A single prompt line such as "Do not remove counts, dates or sample sizes" prevents the most common loss.

Re-checking at 30 days

A plan is not finished until someone checks the evidence. Schedule a short re-check at 30 days: take the actions marked done, ask for the evidence (a record, a signed audit, a changed template) and look at it. For each action, mark it closed with evidence, closed without evidence (reopen it), or overdue with a new date and reason. Then run the same sample again where you found a problem, for example 8 MAR charts on the following four weekends. If the numbers have not moved, the action was wrong or not carried out, and the plan needs a different one. Re-using the same summary prompt on the new findings gives you a quick before-and-after for the team meeting.

Five mistakes a mock inspection makes visible

1. Paper that does not match practice. A care plan says two-hourly repositioning while the charts show gaps. Check practice, not only paper.
2. Evidence that exists but cannot be found. A training certificate sits in a drawer. If you cannot produce it in five minutes, treat it as missing for planning purposes.
3. Policies that nobody follows. Ask two staff to describe the policy in their own words.
4. Actions without owners. The audit lists problems, not who will fix them.
5. AI summaries that smooth the edges. A summary may describe "some inconsistency" where the finding was "a resident waited 15 minutes". Keep the original findings next to the summary and read the originals for anything that touches safety.

The fourth and fifth belong together: a plan with weak wording is easy to agree to and hard to act on. A daily-notes example of the same wording problem is in care home daily notes: better wording with AI, checked by staff.

Presenting the results to the team without blame

How the findings are shared decides whether staff help fix them. Open the team meeting with what was strong, then take the themes, not the individuals. Show the one-page summary, the action table and the date of the 30-day re-check. Ask each owner to say whether the action and the date are realistic; an action nobody believes will not happen.

A practical layout for a 30-minute meeting: five minutes on strengths found, fifteen on the top five risks and their actions, five on questions, five on agreeing owners. Send the action table the same day. Say plainly that the mock is for the home's own learning and that staff who raised concerns during it were helping, not reporting on colleagues. Homes that treat a mock as a hunt for culprits find that the next one gets tidied up in advance, which defeats the point.

One more habit helps: keep a "good practice" list alongside the problems. A carer who sat with a resident to cut food, a cleaner who noticed a loose rail, a nurse whose handover note was clear. Those examples are evidence under caring and well-led, and they are good for morale. AI can draft a short thank-you note for each, which the manager edits and signs.

Keeping an evidence log between mocks

A mock only checks one day. A real assessment draws on what is true across the year, so build a light log that holds one line per month for each key question: what you did, what you found, what changed. At the end of each mock add the plan and the closed actions. After a year you have a story that does not depend on anyone's memory, and that is something to show if CQC asks how you check your own quality.

A short monthly review is more sustainable than an annual scramble. How to run a monthly AI quality review in 30 minutes shows a routine that fits into the same log.

Using the template when you run more than one home

For a provider with several homes, keep the same template and codes so findings can be compared. Ask AI to compare de-identified themes across homes: which key question has the most partial ratings, which findings repeat, which home closes actions fastest. A group quality lead can then spend time where it matters. Keep the home-level detail in each home; send only counts and themes upwards.

What not to hand to AI

• The rating. A mock can say strong, partial or weak on a check; whether the home would be rated good or requires improvement is a judgement for the real assessment.
• Conversations with residents. Staff or the reviewer record those; a transcript tool in a care setting needs consent and a clear lawful basis.
• Safeguarding concerns found during the day. They follow your safeguarding procedure immediately and are not parked in a plan.
• Any statement that a standard is met. Only evidence shows that.

If you are weighing software that promises inspection readiness, care home software: an AI features checklist before you sign lists what to ask. And for a wider rollout beyond this one job, AI for care homes: a registered manager's 90-day implementation plan lays out an order of work. If you would like to talk through the template with someone, that is what my AI implementation consultation covers, at the contact page.

Questions managers ask about mock inspections

Who should carry out a care home mock inspection?

Someone who does not run the area being checked. A manager from another home, a quality lead or a trusted outside adviser sees things the home team has stopped noticing. If it must be done internally, swap areas so nobody marks their own work.

How often should a care home run a mock inspection?

Many providers do a full one once or twice a year and a short focused check on one key question each month. Choose a rhythm you can keep. A smaller check that happens beats a perfect schedule that gets dropped.

Can I paste resident information into an AI tool for the summary?

Not unless your tool and agreement are approved for personal data. Remove names, room numbers and dates of birth before using any general AI tool, and keep findings written as themes such as 'three of ten records lacked a review date'.

Will a mock inspection predict the real rating?

No. It shows where your evidence is thin and where practice may not match your paperwork. A real assessment draws on people's experiences, staff and partner feedback, observation and processes, and a mock can only approximate that mix.

Further reads

• CQC Statement of Purpose Template: Draft Yours With AI and Check It — Keep your statement of purpose accurate before any inspection.
• AI Consultant for a Care Home: Do You Need One, and What to Ask — Whether a care home needs outside help with AI.
• Domiciliary Care Software UK: What the AI Features Actually Do — What AI features in care-visit software actually do.
• Home Care Agency AI: A 90-Day Plan for the Owner — A 90-day plan if you run a home care agency.
• How to Check AI Is Doing Good Work, Not Just Fast Work — How to judge AI output on accuracy, not only speed.
• How to Spot-Check AI Support Replies: A Weekly Sampling Routine — A sampling routine you can borrow for record checks.
• AI Tools for Facility Management in a Care Home: Maintenance Logs — Turn carer messages into maintenance tickets, summarise safety-check logs and chase contractors, with a named person signing off every safety item.
• AI Tools for Nurses Who Run a Team: Rotas, Handover and Audits — A nurse manager's guide to AI for rotas, handover templates and audit write-ups, with a clear list of what must never go into a prompt.
• CQC and GP Practices: Build an Evidence Folder With AI — Organise a GP practice's CQC evidence by key question and quality statement, with an AI-built index, policy review log and mock-inspection summary.
• Care Home Audit Checklist: A Monthly Round Logged and Actioned With AI — A monthly care home audit checklist with scoring, an AI summary built from counts only, and an action log with owners and dates.
• Care Home Mandatory Training Tracker: Renewal Dates and AI Overdue Alerts — A spreadsheet tracker for care home mandatory training with renewal formulas, a weekly overdue list and a manager sign-off step.
• Care Home Medication Audit: A Monthly Template and AI Action Log — A monthly medication audit checklist for care homes: records and counts only, anonymous summaries and a signed action log.
• Care Home Policies and Procedures UK: Keeping Them Current With AI — Use AI to compare each care home policy with current standards, keep tracked changes and a review log, and get staff sign-off before anything is published.
• Care Home Tenders UK: AI Drafting From Your Evidence Library — Build an evidence library once, then let AI draft tender answers from it, with word limits, social value and a quality review that stops invented outcomes.
• Care Inspectorate Self-Evaluation: Draft the Evidence Summary With AI — A Scotland-specific worked self-evaluation: quality framework headings, an evidence summary drafted with AI prompts, and a provider who owns every judgement.
• Home Care Agency Policies and Procedures: A Review-First AI Method — Start from the regulator's standard, draft with AI, and put a named person through a ten-point review before any policy reaches a carer.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: CQC assessment framework, key-question and evidence-category pages, and CQC consultation pages on the 2026 draft sector-specific frameworks (archived copies from June to October 2026), read November 2026. Check the CQC website for the framework in force on the day.