The clearest red flags are a price that jumps after the first year, billing based on contacts you've lost, accounts or lists held in the provider's name, AI that posts or replies with no approval step, and guarantees about rankings or AI-search placement. Any one is worth questioning; two or more usually means walk away.
Most of these aren't scams. They're sales habits that make a quote look cheaper or a product look cleverer than it is, and they cost small businesses money quietly, often in year two. Each flag below comes with what it looks like on a quote or in a demo, and how to check it in a few minutes.
Red flags in the price
1. A first-year price that doubles on renewal
What it looks like: "50% off for 12 months" on an email platform, or a CRM starter tier at a promotional rate that renews at list price. Both are common, and both are fine if you budget for year two. They become a problem when the business case was built on the year-one number.
How to check: ask for the renewal price in writing and build your comparison on it. HubSpot, for example, often runs steep first-year promotions on its Starter tier; its list price is $20 a seat a month, and that's the number to plan around.
2. Paying for contacts you've already lost
What it looks like: a contact-based price tier that you keep outgrowing even though your active list hasn't grown. Mailchimp's own help pages say subscribed, unsubscribed and non-subscribed contacts all count towards the plan's contact limit; only cleaned, archived, pending and deleted contacts don't.
Example: an illustrative butcher with a meat-box email list of 4,200 addresses, 900 of them unsubscribed over three years, is paying for a tier sized for 4,200 while mailing 3,300. Archiving the unsubscribed contacts can drop a tier.
How to check: ask exactly which contacts are billable, and whether archiving or deleting them changes the bill. Then look at your own list's status breakdown.
3. Usage-based AI with no cap and a loose definition
What it looks like: "Only pay for results" chatbot pricing, AI credits, or per-message charges. Per-use pricing can be fair; the danger is not knowing what counts. Intercom, for instance, charges from $0.99 per Fin "outcome" and defines an outcome as a customer confirming the issue is resolved, not asking for more help, or Fin completing a workflow. That middle condition means a customer who simply gives up can count as an outcome.
How to check: ask for the exact definition, a monthly cap or alert, and last month's usage from a similar customer. Then model the bill at three times your expected volume.
Red flags in the contract and the accounts
4. A 12-month lock-in before you've seen any results
What it looks like: an agency retainer with a 12-month minimum and 90 days' notice, or software that only offers annual billing with auto-renewal. Your first three months are when you find out whether it works; a contract that prevents acting on that finding is a red flag.
How to check: ask for a monthly option or a three-month break clause, even at a higher price. Put the renewal date and the last day you can give notice in your calendar the day you sign. What to look for in auto-renewals, price rises and notice periods covers the wording.
5. The provider owns your accounts, list or domain
What it looks like: ads run from the agency's ad account, your email list held in their platform, a chatbot's contacts stored in their tool, or a domain registered in their name. It's usually presented as convenience.
Example: an illustrative farm shop leaves an agency after two years and discovers the Google Ads history, the audiences and the conversion setup all lived in the agency's account. It starts from zero, and the new campaigns spend their first weeks relearning what the old ones knew.
How to check: for every account, ask "whose name is it in, and who is the admin?" The answer should be your business, with the provider as a user you can remove.
6. No clean way to export what you've built
What it looks like: "Your content lives in our platform." Posts, email templates, chatbot answers and prompt libraries that can't be downloaded in a usable format.
How to check: during the trial, try exporting your contacts, one email template and the chatbot's knowledge base. If you can't, assume you'll lose them. Keeping your data portable from the start is covered in how to avoid AI vendor lock-in.
Red flags in what the AI is claimed to do
7. Guaranteed rankings, or guaranteed placement in AI answers
What it looks like: "Guaranteed page one", "we'll get you recommended by ChatGPT", or an "AI search optimisation" package built around adding an llms.txt file and special markup.
Why it's a flag: nobody controls those results. Google's own guidance says its AI Overviews and AI Mode need no special optimisation beyond normal search good practice, and that llms.txt files and special markup aren't used for them. A package priced around those tricks is selling reassurance.
How to check: ask what specifically they'll change on your site and listings, and what they'll measure. Real work sounds like fixing your Google Business Profile, clearer service pages and more genuine reviews, not secret techniques.
8. "Fully autonomous" posting or replying, with no approval step
What it looks like: a tool or service that writes and publishes social posts, emails or customer replies without a human seeing them first. Sold as time saved.
Example: an illustrative delicatessen's auto-generated post describes a new cheese as "perfect for pregnant customers" because the tool guessed from the word "pasteurised". Nobody read it before it went live. One approval click would have caught it.
How to check: ask where the approval step is and whether it can be made compulsory. For food businesses, anything touching allergens, health or dietary claims should never publish unreviewed.
9. The demo only ever runs on their sample data
What it looks like: polished results on a fictional clothing brand, and reluctance to try your real product list, your customer questions or your past emails before you sign.
How to check: insist on a trial with your own material. Give the chatbot ten real questions from your inbox. Ask the writing tool for a product description of your least typical product. How it handles your oddities is the whole test.
10. "Proprietary AI" that is a markup on a chat subscription
What it looks like: $199 a month for a "marketing AI" that, when you look closely, is a set of prompts running on a mainstream model. Sometimes that's worth paying for, because the prompts and workflow are good. Often it isn't.
How to check: ask which underlying model it uses and what it does that ChatGPT Plus or Claude Pro (each $20 a month) plus your existing email tool can't. If the honest answer is "saves you writing the prompts", compare that with an afternoon spent writing your own. What's worth paying for among free and paid AI marketing tools helps with that comparison.
11. Changes switched on in your ad account without telling you
What it looks like: budgets that creep up, new keywords you didn't choose, broad match appearing across campaigns. Google Ads has an auto-apply recommendations setting that can make certain changes automatically; some tools and agencies switch it on.
How to check: in Google Ads, the Recommendations page has auto-apply settings with a History tab showing what was applied and when. Ask any provider to tell you in writing before enabling it.
Red flags in data and ethics
12. Vague answers about where your customer data goes
What it looks like: "It's all secure" in reply to "which AI tools will see our customer list, and do they train on it?"
How to check: ask for the list of tools, the plan each is on, whether each trains on your data by default, and whether they'll sign a data processing agreement. Business tiers of the main AI assistants don't train on business content by default; consumer tiers depend on a privacy switch. A provider that can't answer this can't protect your customers.
13. Features that manufacture social proof
What it looks like: AI-written reviews or testimonials, "review generation" that only invites happy customers to Google while steering unhappy ones to a private form, bought followers, or engagement pods.
Why it's a flag: Google's review policies ban fake engagement, including reviews not based on a real experience and reviews paid for directly or in kind, and Google can restrict a Business Profile that breaks them. The private-form trick is known as review gating; what review gating is and why it can get you penalised explains the risk. Consumer-protection law in many places also treats fake reviews as misleading. Asking every real customer for an honest review is fine and works.
14. A product built on something that's being switched off
What it looks like: a service whose "custom AI assistant" is a custom GPT, or a tool that depends on one small supplier with no plan B.
Why it's a flag: OpenAI is retiring custom GPTs, which stop running on 11 December 2026. When the AI calendar tool Clockwise shut down in March 2026 it deleted user data rather than transferring it. Suppliers disappear, and small AI suppliers disappear faster.
How to check: ask what the product is built on, what happens to your data if the company closes, and whether you can export everything now. Then do the export once, just to prove you can.
15. Terms that let the vendor change your data settings quietly
What it looks like: a clause saying the vendor may update its privacy practices or product settings "from time to time", with notice given only by updating a web page. Settings that were off when you signed can end up on.
A real example of the pattern: since 16 June 2026, new users of SimplePractice's Note Taker, a clinical note tool, have been opted in by default to the vendor keeping de-identified transcripts. That's outside marketing, but the lesson carries over: defaults are the vendor's choice, and they change.
How to check: ask whether you'll be emailed before any change to how your data is stored or used, and whether existing settings are kept when defaults change. Then put a quarterly reminder in your calendar to look at the privacy and data settings of each marketing tool you pay for. It takes ten minutes.
A delicatessen's quote, marked up
Here is an illustrative quote a delicatessen with an online hamper shop received for an "AI marketing suite", with the flags marked.
| Line on the quote | Year-one cost | Flag | What to ask for |
|---|---|---|---|
| Platform licence, "50% launch discount" | 12 × $149 = $1,788 | 1: renews at $298 | Year-two price in writing |
| Contacts tier up to 5,000 (list is 5,100 incl. 1,400 unsubscribed) | Included | 2: billed on lost contacts | Which statuses are billable |
| AI chatbot, "pay per result", $0.90 | Estimate $540 | 3: no cap, loose definition | Definition of a result, monthly cap |
| Minimum term 12 months, 90 days' notice | n/a | 4 | Three-month break clause |
| Ads managed in "our partner account" | Management $2,400 | 5 | Account in the deli's name |
| "AI search visibility boost" | $600 | 7 | Remove, or specify changes and measures |
| Auto-posting, 30 posts a month | Included | 8: no approval step | Compulsory approval before publishing |
| Total | $5,328 |
Year two, with the licence at full price and nothing else changed, rises to $7,116. The owner took the quote back with four requests: an account in her own name, an approval step, a cap on chatbot charges and the removal of the "visibility boost". Two were accepted, two weren't, and she bought a smaller package from a different provider using the same list of asks. The marked-up table did the negotiating for her.
A ten-minute screening call that surfaces most of these
Ask these on the first call and note how directly each is answered. Hesitation on questions 2 and 5 matters more than anything else.
1. What will the price be in month 13, in writing?
2. Whose name will each account and list be in, and who is the admin?
3. What exactly counts as a billable contact, credit, message or result?
4. Can I pay monthly for the first three months?
5. Who approves posts, emails and replies before they go live?
6. Which AI model and which plan does the product run on?
7. Which of our customer data goes where, and is any of it used for training?
8. Can I export contacts, content and chatbot answers today, during the trial?
9. What happens to our data if your company closes?
10. Will you promise rankings, reviews or AI placement? (The right answer is no.)A good provider answers most of these in a sentence and sends the written version without being chased. If you want the longer list for comparing full proposals, the questions to ask before you pay anyone to set up AI marketing goes into each area in more depth.
Spotted a flag after signing? What to do next
Most owners find these after the contract has started, usually when the first renewal invoice arrives. It's rarely too late to improve the deal, as long as you act in the right order.
- Secure what's yours first. Export contacts, content, reports and chatbot answers today, before raising anything. If an account is in the provider's name, ask for admin access in your business's name as a routine request, not a complaint.
- Find the dates. Renewal date, notice deadline, any price-review clause. Put them in your calendar with a two-week warning.
- Raise one issue at a time, in writing. A clear, specific request ("please confirm the ad account will be transferred to us") gets a better response than a list of grievances.
- Offer something in return if you need to. Providers will often trade a longer commitment for a lower price or better terms, once you know the product works.
- Get advice before refusing to pay. If the dispute is about money or a contract term, talk to a solicitor or your business adviser before withholding anything.
An illustrative butcher found flag 2 in month eight: his email platform had moved him up a contact tier, and 1,100 of the contacts it was counting had unsubscribed years earlier. He archived them, dropped back a tier, and then emailed the platform's support team asking whether the two months billed at the higher tier could be credited. They credited one. Total time spent: about forty minutes. Total saved over a year: several hundred dollars, and the list he mailed was more accurate afterwards too.
A farm shop that found flag 5 was less lucky: its agency held the ad account and wouldn't transfer it. The shop gave notice, opened a fresh account in its own name, and rebuilt the two campaigns in an afternoon from screenshots it had taken of the old settings. Results dipped for about three weeks while the new campaigns learnt. The lesson it wrote into its supplier checklist afterwards: screenshot every campaign's settings quarterly, whoever runs them.
More questions about spotting a bad AI marketing deal
Is a free trial enough to judge AI marketing software?
Only if you use it on your own material. Load your real product list, write three real emails or posts and run your own customer questions through any chatbot. A trial that stays on the vendor's sample data tells you how the demo behaves, not how the tool will behave with your stock names, prices and customers.
Are annual contracts always a bad idea?
No. Annual billing usually saves money once you know the tool works for you. The problem is committing to a year before you have that evidence. Pay monthly for the first two or three months, then switch to annual if the numbers hold up, and diary the renewal date and notice period the day you sign.
What should I do if I've already signed something with these red flags?
Start with ownership. Export your lists, content and reports now, ask for admin access to any account in your business's name, and write down the notice date. Then raise the specific issue in writing. Many providers will fix ownership or billing problems when asked plainly, before it becomes a dispute.
Is it a red flag if a provider uses ChatGPT to do the work?
Not in itself. Plenty of good freelancers use general AI assistants with skill and judgement. It becomes a red flag when the tool is presented as proprietary technology to justify a high fee, or when your customer data goes into a consumer account with training switched on. Ask what they use and on which plan.
Further reads
- AI Consultant Red Flags: 12 Warning Signs to Walk Away From — Warning signs specific to hiring an AI consultant.
- Hidden Costs of AI Implementation Most Small Businesses Miss — The costs beyond the subscription that catch small businesses out.
- What If Your AI Vendor Shuts Down? Checks Before You Commit — Checks to make before relying on a small AI supplier.
- How to Evaluate an AI Software Vendor: A Small Business Scorecard — A scorecard for comparing AI software vendors side by side.
- Can You Use AI to Write Testimonials? Where the Legal Line Is — Where AI-written testimonials cross the legal line.
- Annual or Monthly Billing for AI Tools: Which Saves You More? — When switching to annual billing actually saves you money.
- Can AI Manage Your Google Ads Without an Agency? — What Google's own AI handles, the jobs it leaves with you, a catering company's move from agency to DIY and a 45-minute weekly routine.
- DIY or Done-With-You: Who Should Set Up Your AI Marketing? — A scoring table for choosing DIY, done-with-you or done-for-you AI marketing, the setup jobs ranked by difficulty, and what a proper handover includes.
- AI Review Management Tools for Small Businesses (2026) — Which review tools are worth paying for: verified prices, what their AI actually does, and picks for a garage, an online supplier and a multi-branch agency.
- Best AI Landing Page Builders for Small Businesses (2026) — Verified prices, traffic caps and AI features for the landing page builders small businesses actually use, with picks for three different businesses.
- AI Social Listening: Tracking What Customers Say About You — Build a small listening process that separates real customer themes from duplicate posts, wrong matches and misleading scores.
- AI Washing: How to Spot Software That Oversells Its AI — A grouped checklist for testing software AI claims, from sales-page wording to live demos and accuracy numbers, with a scored worked example.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Mailchimp help pages (which contact statuses count towards plan limits) and pricing page; Intercom pricing page (Fin outcome definition); Google Ads Help (auto-apply recommendations); Google Business Profile prohibited content policy; OpenAI help pages (custom GPT retirement).