DDS Web Solutions
AI & Automation

Evaluate an AI marketing tool before you buy: one written job, a short trial on your own work, and no unsourced promises

9 min

The short answer

Write one job the tool should do, and one way you will check it. Record a baseline for a recent period. Run a short trial on your own pages, ads, or reviews, not on the vendor's sample data.

Keep names, phone numbers, and visit reasons out of the tool. Have a person fact-check every public draft, including titles and descriptions, before anything is published.

Do not buy on a percentage the vendor cannot source. This is marketing and web guidance. It is not legal advice.

Claims that need proof

A vendor may lead with a lift in appointments, leads, or time saved. Those numbers only mean something if they name what was measured, over what period, against what, and for what kind of practice.

Ask three more questions. Does the tool do the job better than what you do now, and how was that shown? Is the AI part doing the work, or is it a template with an AI label? What can go wrong, and what does the vendor do when it does?

A simple test: if you would not put the vendor's claim on your own site, do not use it to justify the purchase. How dental advertising claims are handled more broadly is in FTC guidelines for dental and medical advertising.

Patient details stay out

Marketing drafts do not need a patient's name, phone number, or the reason for a visit. Keep those out of prompts, uploads, and chat logs.

Ask the vendor, in writing, whether your input is used to train a model for other customers, where the data is stored, and how you can delete it if you cancel. Limit access to named people at the practice, and remove people who leave.

Whether a given tool may see patient information, and which agreements you would need, are questions for the practice's own counsel. The same caution for analytics events is in tracking calls, forms, and bookings without patient details.

If the tool writes office pages, blog posts, or metadata, read Google's guidance before you turn it loose. Search Central says generative AI can help with research and with giving original work a structure. It also says using such tools to generate many pages without adding value for users may violate the spam policy on scaled content abuse.

Google also notes that generative models predict likely words; they do not look up facts. Outputs can contain inaccuracies, sometimes called hallucinations. Every draft should be checked by a person for accuracy before it is published. That review includes title tags, meta descriptions, structured data, and image alt text, because those can appear in search results.

Sharing how a piece of content was created can give readers useful context. Whether you must say so in a given case is a legal question, not a ranking tip. A related playbook for groups is AI search and AI Overviews for multi-location dental groups.

A short, written trial

Before

Write the job, the check, the cost, and a go or no-go rule. Record a baseline for a recent period, using the same measure you will use in the trial.

During

Use your own work, not sample data. Involve the people who will actually use the tool. Keep a log of time spent, drafts published, and anything that had to be rewritten.

After

Compare the log to the baseline and to the written rule. If the tool did not do the job, cancel. If it did, write who owns it, who reviews drafts, and how leavers lose access.

NIST's AI Risk Management Framework is voluntary. It is a way to think about risks to people and organizations when you design, buy, or use an AI tool, not a score you have to hit.

Red flags

A lift number with no baseline

A claim such as more appointments or more leads has to name what was measured, over what period, and against what. If the vendor cannot show that, treat the number as advertising, not evidence.

A demo only, with no trial on your work

A polished walkthrough on sample data does not show how the tool behaves on your pages, ads, or reviews.

Your patient list has to leave the practice

If setup means uploading names, phone numbers, or visit reasons, stop and talk to counsel before going further.

No written answer on training data

Ask whether your prompts, pages, or patient-facing text are used to train a model for other customers, and get the answer in writing.

No way to export what you created

If you leave, you should be able to take your drafts, prompts, and settings in a standard format.

It is sold as AI-powered, but the work is a template

Ask what the AI part does in the product you would use. If the answer is vague, judge it as the template it may be.

Checklist

AI tool trial checklist

0 of 12 done

A practice-level look at earning mentions in AI answers is in earning citations in Google AI Overviews. How a group's offices show up in AI answers is in AI search for multi-location dental groups.

For a dental practice or group

DDS Web Solutions builds websites and runs search, ads, and reporting for dental practices. Our team has been building websites since 2014, we have served 100+ practices, and our offices are in Sacramento and San Mateo.

We can help you write the job for a tool, keep patient details out of marketing systems, and review public drafts. We do not sell AI tools, and we do not rank vendors. See AI marketing services, or request a quote.

FAQ

How should a dental practice judge an AI marketing tool?

Write one job and one way to check it, record a baseline, and run a short trial on your own work. Keep patient details out. Have a person review every public draft. Do not buy on a percentage the vendor cannot source.

Can we use an AI tool with patient information?

A practical rule for marketing work is to keep names, phone numbers, and visit reasons out of prompts and uploads. Ask counsel before any tool would see that kind of information.

Will AI-written pages help us in Google Search?

Google's guidance is that generative AI can help with research and structure, but pages created at scale without added value may violate its spam policy on scaled content abuse. It also says generative models can invent facts, so every draft and every title, description, and alt text should be checked by a person before it is published.

Do we have to tell patients when we use AI to write?

Google suggests that sharing how a piece of content was created can give readers useful context. Whether a practice must disclose that in a given case is a legal question for counsel.

What if a vendor says the tool will raise appointments by a set percent?

Ask what was measured, over what period, against what baseline, and for what kind of practice. If the vendor cannot show that, set the number aside and judge the tool on your own trial.

Should we build our own tool instead of buying one?

For most practices, buying a tool and writing a short trial is the simpler path. Building something in-house is a project with its own cost, maintenance, and review.

Sources

Explore Our Services

Next step

Need help turning reading into results?

Our team specializes in dental and healthcare marketing. Get a quote and we will scope a plan that fits your practice.