What does a code audit of my AI-built app include, and what should it cost?

By Ranjit Bhinge, founder of Blur Studio · Updated · 8 min read

Short answer: A good audit of an AI-built app (one made with a tool like Lovable, Bolt or Replit, where you describe the app and the AI writes the code) is a written report that tells you who can see your data, whether your passwords and keys are exposed, and whether the app can be safely launched, kept running and handed to someone else. As of November 2026, published prices run from about $199 for a fixed-price report at one vendor, to about 1,500 pounds at another, to $5,000 and up for a full penetration test (a hired attack on your app to see what breaks). Pay for the smallest review that matches how risky your app is.

Ranjit Bhinge

You built an app with AI, real people are about to use it, and you want to know what a proper review covers and what a fair price looks like before you give anyone access to it.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

A code audit sounds scary and expensive. It does not have to be either. For most small business apps, an audit is a few days of an engineer reading your app and writing down what is wrong, in order of how bad it is. You then decide what to fix.

The problem is that almost everyone selling audits is also selling the fix. So the scope, the price and the wording of the report all lean toward more work. This page gives you a neutral view: what a sensible audit covers, what the findings look like, what published prices say, and how to read a quote.

What an audit is and is not

An audit is a review. Someone reads your code and your settings, tries a few things, and reports what they found. It is not a rebuild, and it does not fix anything unless you pay for that separately.

There are three things people call an audit, and they are different jobs:

TypeWhat happensBest for
Automated scanA tool checks your app for known problems. Lovable, a tool that builds apps from plain-English prompts, has built-in scans. Its docs say they cannot guarantee complete security and do not replace a thorough security review.A first pass, any time, at no extra cost
Code and settings reviewAn engineer reads the code, the database rules and the hosting setup, then writes a report ranked by severity.Most small business apps before launch
Penetration testA security specialist plays the attacker against the running app and reports what they broke into.Apps holding payments, health or financial records, or sold to other businesses

Lovable's own docs say that for apps that handle sensitive data or critical functions, you should add a professional security review. That is the gap an audit fills.

What a good audit includes

Ask for these six areas in writing. If a quote skips one, ask why. Our own AI-built app security checklist goes deeper on each, and the production-ready guide covers what "ready" means beyond security.

  1. Who can see and change your data. Most AI-built apps keep data in Supabase, a hosted database service. Supabase's docs say a table without row-level security (rules that say which user can read or change which row) is readable and writable by any role with access to it. The reviewer should test this with a real logged-out visitor and a real second user.
  2. Exposed keys and passwords. Secret keys must never sit in code that runs in the browser. Supabase's docs say its secret keys bypass the rules entirely, so one in the browser is a full leak.
  3. Login and permissions. Can a normal user reach an admin page by typing its address? OWASP, the nonprofit that publishes the standard list of web app risks, puts Broken Access Control first on its 2025 Top 10.
  4. Libraries and code you did not write. Your app pulls in dozens of packages. The reviewer checks for known problems in them. OWASP's 2025 list has a category for exactly this: Software Supply Chain Failures.
  5. Payments and personal data. Where is customer data stored, who can export it, and does the app log things it should not?
  6. Backups, ownership and recovery. Can you get your code and data out? Is there a backup that has been tested? Who owns the hosting account? These are boring and they decide whether a bad day is a small one.

A review of the app's structure is also worth asking for if you plan to grow it: is it a pile of patches, or something a developer could keep building on? That is the question behind whether AI can finish a half-built app or you need a developer.

What findings look like in plain English

A report should read like a list a non-technical owner can act on. Here are four made-up examples of the form it should take. They are illustrations, not findings from a real app.

SeverityFindingWhat it means for youFix effort (example)
CriticalThe customers table has no access rules.Anyone who finds your public key can download every customer record.Hours
HighA secret key is inside the browser code.Anyone can use it to act as your app. The key must be replaced.Hours to a day
MediumAny logged-in user can open the admin page by its address.A customer could change prices or see other customers.A day
LowNo backup has ever been restored as a test.If something breaks you may not be able to recover.Half a day

Good reports also say how the reviewer found the problem, so a developer can repeat it, and give a recommendation: keep, fix or rebuild. SpotDev, a UK studio that sells audits of AI-built apps, lists exactly that "keep, fix or rebuild" call in its deliverables.

What audits cost: published prices

There is no reliable average, so here is what vendors publish, as of November 2026.

OptionPublished priceWhat it includes
Valletta Software, a software firm, audit reportFrom $199, fixed priceEight review areas scored, a risk map ranked by severity, a fix roadmap, a fixed-price quote for the cleanup and a debrief call.
SpotDev, audit of one appFrom 1,500 pounds, excluding VATA written report graded by severity, a walk-through with the engineers, and a keep, fix or rebuild recommendation. The exact price depends on a risk tier, from an internal tool with no personal data up to software you sell to others.
Wavect, a software firm that does vibe code rescue workNo price publishedIts cost guide says an initial audit often takes a few focused days to about a week, and that scope is set after an initial look.
Full penetration testBright Defense's guide puts web app tests at $5,000 to $30,000A specialist attacks the live app. Bright Defense, a security firm, also lists its own fixed-scope plans from $2,750 to $9,250, and says a standalone retest can cost around $2,000 to $5,000 if it is not included.

Two cautions. First, the "from" in a price is the cheapest tier, usually for the simplest app. Second, the penetration test figures come from a security vendor's own guide, so they lean toward what that vendor sells. Get a few scoped quotes before you plan a budget.

What moves the price: how many kinds of users the app has (customer, staff, admin), how many connections to other systems, whether it takes payments, and whether it holds sensitive data. Bright Defense's guide lists user roles, pages, integrations and API testing as the main drivers.

Worked example with made-up numbers

Say you run a 12-person training company and built a booking portal with Lovable. Customers log in, see their bookings and pay a deposit. It holds names, emails and phone numbers.

  • It is customer-facing and holds personal data, so it is a middle-risk app. A full penetration test is probably more than you need today.
  • A code and settings review fits. Budget a low four-figure sum using the published ranges above as a guide, and ask for a fixed price.
  • Say the report finds one critical item (a table with no access rules), two high items and some low ones. Total fix time is about three days of developer work.
  • You fix the critical and high items before launch. You write the low items into a list for later.

These figures are an illustration. Your app and quotes will differ.

Red flags in an audit quote

  • No written scope. You should be able to read exactly which areas are covered and which are not.
  • The audit is free but the fix is open-ended. A free audit that ends in "we will bill hourly to fix it" is a sales call. Wavect's guide says you should not be asked to sign a fixed figure for work nobody has looked at yet, and I would add that you should not sign an open-ended one either.
  • They want your admin logins on day one. A reviewer needs read access to the code and a test account. They rarely need owner access to your hosting or payments.
  • Scary findings with no explanation. A report that says "critical" and does not say how to repeat the problem is padding.
  • A guarantee of security. Nobody can promise that. Even Lovable says its scans cannot guarantee complete security.
  • No retest. Ask whether checking your fixes is included or priced separately.

What to ask before you hire anyone

  1. Which of the six areas above are covered, in writing?
  2. What access do you need, and can it be read-only?
  3. Is it fixed price? What happens if the app is bigger than it looked?
  4. Do I get a report I can hand to another developer or an investor?
  5. Is a retest included?
  6. If the fix is a rebuild, will you tell me even though it ends your fee?

If you are unsure how an audit fits into getting your app live, see how to move an AI-built app to production and, if the app was built on Lovable, how to hire a Lovable developer.

What you can do yourself before paying

  • Run the built-in scan in your builder. Fix what it flags.
  • Open your app in a private browser window, logged out, and try to reach pages that should need a login.
  • Make two test accounts. Log in as one and see whether you can view the other's data.
  • Search your code for words like "secret", "key" and "password" and check none are written in plain text.
  • Confirm you can export your database and download your code today.

These checks do not replace an audit. They make an audit cheaper because the reviewer starts further along. And they are a good test of whether you need one.

Who should not pay for one yet

If your app is a private tool used by you and two colleagues, with no customer data, the built-in scan plus the checks above is probably enough for now. If your app is still changing every day, wait until it settles, or the report will be out of date in a week.

How I'd approach this

I would match the review to the risk. A private tool gets the scan and the self-checks. A customer-facing app with personal data gets a scoped code and settings review at a fixed price. Anything with payments, health records or business customers who will ask for proof gets that review and then a penetration test.

I would also insist on one rule: the person who audits is paid for the report, and the fix is a separate decision. That keeps the report honest. I run a Go-Live Plan Audit for owners who want exactly this, a written plan before launch. If a report finds problems you would rather have someone fix, Blur Studio fixes and finishes start around $1,500. You can also take the report to anyone you like.

Questions people ask

How long does an audit of an AI-built app take?

Published guidance from vendors puts a first review at a few focused days to about a week for a bounded app, as Wavect's cost guide says. Apps with payments or sensitive data take longer, and SpotDev says its timeline depends on the risk tier and app size. Ask for a delivery date with the price.

Is the free security scan inside my app builder enough?

It is a good first pass, not a full review. Lovable's docs say its scans cannot guarantee complete security and do not replace a thorough review, and they recommend a professional review for apps that handle sensitive data. Run the scan, fix what it flags, then decide whether your app's risk justifies a paid audit.

Do I need a penetration test or just a code review?

For most small business apps a code and settings review is the right first step. A penetration test, a hired attack on the live app, makes more sense when the app handles payments, health or financial records, or you sell it to other businesses. One security vendor's guide, Bright Defense, puts web app tests at $5,000 to $30,000 as of November 2026.

What access does an auditor need?

Read access to the code and a test login for each type of user. They should not need owner access to your hosting or payment accounts. Give the least access that lets them do the job, and remove it afterwards.

Will an audit tell me whether to keep or rebuild the app?

A good one should. Ask for a keep, fix or rebuild recommendation in writing. If the app is half finished, read whether AI can finish it or you need a developer as well.

Can I use an audit report to reassure customers or investors?

It helps, but it is not a certification. Customers and investors will usually ask for the report summary and proof that the serious findings were fixed. Lovable's own trust page is described in its docs as informational and not a certification, and the same logic applies to a one-off audit.

More in this topic: Problem solving: AI-built and no-code apps (26 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…