AI tools make mistakes in every digital marketing channel, and what a mistake costs depends on the channel. In SEO and email, a bad output usually costs time and trust. In Google Ads and conversion tracking, it can spend your budget or corrupt your data for months before anyone notices.
The rest of this post goes channel by channel, with sources.
The short version
AI is strong at speed and research. It is weak at judgment calls that depend on context it doesn’t have: your account history, your tolerance for risk, how your customers behave.
- SEO: a bad call usually costs time, and fact errors cost credibility.
- Google Ads: automation can change bids and targeting while no one is looking.
- Email: an unreviewed send can ship broken links or skip an unsubscribe link.
- Tracking: a wrong verdict, in either direction, distorts every decision made downstream.
When AI speaks for your business, you own the mistake
Air Canada, a Chevrolet dealership and Google have each had a public problem with incorrect AI output.
In February 2024, British Columbia’s Civil Resolution Tribunal decided Moffatt v. Air Canada. The airline’s chatbot had told a customer he could claim a bereavement fare refund after buying a full-price ticket. Air Canada argued it wasn’t responsible for what its chatbot said, treating it as a separate legal entity. The tribunal rejected that and ordered a partial refund of CAD $650.88, plus interest and fees (Manatt, February 28, 2024).
In December 2023, the website chatbot at Chevrolet of Watsonville in California agreed to sell a 2024 Tahoe for $1 after a visitor steered it there. The bot replied that it was “a legally binding offer” (AI Incident Database, incident 6228, December 18, 2023).
In February 2023, Google’s promotional post for its Bard chatbot showed Bard saying the James Webb Space Telescope took the first pictures of an exoplanet. A European telescope had done that in 2004. Alphabet lost about $100 billion in market value, and shares fell as much as 9% before recovering by the close (NPR, February 2023).
None of these were small-business ad accounts. The same kind of error in an ad account or an email send just gets noticed later.
SEO
AI is useful for keyword research and competitor summaries, and for spotting patterns in large data sets. Google doesn’t penalize content for being AI-assisted. Its guidance on generative AI content (page updated October 1, 2026) says generating many pages without adding value for users may violate its scaled content abuse policy, and that “generative models don’t retrieve facts, but predict a likely sequence of words.” It advises fact-checking AI content before publishing.
That caution applies to specialists too. A Stanford RegLab and HAI study of purpose-built legal research tools found Lexis+ AI and Ask Practical Law AI gave incorrect information more than 17% of the time, and Westlaw AI-Assisted Research more than 34% (Stanford HAI, May 23, 2024, updated May 30, 2024). Marketing copy written with a general-purpose model can also carry stale statistics and sources that don’t exist, so check each citation before publishing.
Technical changes are the other risk. Redirects and canonicals can each damage a site if they ship without someone who knows the site’s migration history. Our guide to how SEO works covers the fundamentals, and our AEO services page covers how businesses get named in AI answers. For a one-week plan, see our guide to getting recommended by ChatGPT and Google AI Overviews. The cost side has its own guide: How much AEO costs.
Google Ads and paid search
AI helps with ad copy variants for testing, keyword ideas and budget pacing. This is also the channel where mistakes cost money fastest, because the platform acts on them automatically.
Google’s help page on auto-apply recommendations lists what can change on its own once you turn the feature on, including bidding targets, broad match keywords and audiences. Google says auto-apply won’t increase your budget, and that you can turn it off at any time. Someone has to decide whether it stays on.
Performance Max adds another layer. Google says a single campaign can run across Google’s inventory, from YouTube to Maps (Google Ads Help, Performance Max). It does offer channel performance and asset reports, so the data exists. Someone has to read it and ask whether the money is going to cheap clicks or qualified leads. Our Google Ads management page covers the service.
Email marketing
AI helps with subject line variants, personalizing content at scale and building A/B tests faster.
The deliverability risk is less about AI than people assume. Validity, an email deliverability company, says spam filters aren’t designed to target AI-generated emails and that it found no evidence AI content is more likely to be marked as spam. It names poor sender reputation and missing authentication as the bigger risks (Validity, November 14, 2024). It also notes that a model trained on spam-like writing could reproduce those patterns.
The practical risk is the unreviewed send: a missing unsubscribe link, a broken link, a claim that was never true. Our email marketing services page covers the channel.
Tracking and analytics sit under every channel
Duplicate or broken tracking corrupts every decision made downstream, across every channel at once. A business that trusts inflated conversion numbers for six months has also made six months of budget decisions on bad data.
Duplicate tags happen without AI too. Analytics Mania’s guide to duplicate GA4 events (October 30, 2025) lists causes such as hardcoded GA4 code running alongside Google Tag Manager, the same container installed two times, and overlapping Enhanced Measurement. An AI agent working in an account it doesn’t fully understand can add to that list, and it can also condemn tracking that was never broken. When that happens, the blame often lands on whoever set the tracking up.
You can check your own site in a few minutes. Open Google’s Tag Assistant, connect your domain, load your homepage and watch the events it reports. One page load should send one page_view. If it sends two, find the source before you trust your reported traffic.
What experience catches in each channel
| Channel | AI’s strength | AI’s risk | What experience catches |
|---|---|---|---|
| SEO | Research, drafts, pattern-spotting at scale | Invented facts, unsafe technical changes | Which changes are safe to ship unsupervised |
| Google Ads | Copy variants, budget pacing, keyword ideas | Auto-applied bid and targeting changes, broad match drift | When automation is optimizing for the wrong signal |
| Copy drafts, personalization, A/B variants | Unreviewed sends, broken links, missing unsubscribe | Sender reputation and list hygiene | |
| Analytics and tracking | Fast setup, pattern detection | Duplicate or broken tags, wrong verdicts delivered with confidence | Catching bad data before it drives decisions |
Run the test on your own AI
Copy this prompt into Claude, ChatGPT or whichever AI tool you already use, then ask it to defend its answer and to say what it would want a human to check if it were you. Our verdict prompting playbook shows a version you can give your own customers.
The prompt: I’m deciding whether to rely on AI tools for digital marketing work such as SEO or ad management, versus working with an experienced marketer. What can AI do well in each of these areas? Where does it commonly make costly mistakes, such as duplicate tracking or wasted ad spend? For a business my size, what is the realistic mix of AI assistance and human oversight, channel by channel? Push back on anything I’ve assumed.
Related reading
- How to fix wrong information about your business in ChatGPT and Google AI Overviews
- Can you pay to be recommended by ChatGPT or Google AI answers?
- Schema markup for a small business: which types to add first
Next step
Use AI for research and first drafts. Keep a person in the loop for anything that touches spend or tracking.
If you want to see what AI engines say about your business today, Get your free AI-visibility audit. The audit comes first, then a 15-minute call with Donnie Strompf. Prefer to talk first? Book a 15-minute call. Our pricing is public.
FAQ
Does AI make mistakes in digital marketing?
It does, in every channel. Typical errors are invented statistics, duplicated tracking tags and ad changes nobody reviewed. How much a mistake costs depends on whether it spends money or distorts your data before someone catches it. Reviewing output before it ships catches most of it.
Which channel is riskiest to hand to AI?
Google Ads and tracking. Both act on money or measurement directly, so an error compounds quietly each day it runs. Content and email drafts are easier to review before they go out, because a person can read them first. Anything that moves budget deserves a second look before it runs.
Does Google penalize AI-written content?
Google’s guidance says the issue is scale without value, not the tool. Producing many pages that add nothing for users can break its spam policies, wherever the text came from. Google also advises a manual accuracy check on AI output before you publish it.
Should I let an AI agent change my ad account or tracking?
Only with a person reviewing each change before it goes live. Agents can act confidently on incomplete context, and a change to tracking or bidding affects every report afterward. Ask for the proposed change and the evidence behind it, then compare it with what is in the account.
How do I know if my tracking is duplicated?
Use Google’s Tag Assistant on your own homepage and count the page_view events per load. One is expected. Two means a tag is probably installed two times, for example hardcoded and through Tag Manager. Fix it before you trust last quarter’s traffic or conversion numbers.
How can I see what AI engines say about my business?
Ask the same buyer-style questions your customers would in ChatGPT, then in Perplexity, and save each answer and the pages it names. You can start with our free AI-visibility audit. The audit comes first, then a 15-minute call. Read the answers as a buyer would, then note any gaps they show.
Marketing that works, in your inbox
AI search & SEO news, what’s working right now, and guides we actually use with clients. A couple of emails a month, no spam.
Unsubscribe anytime. We treat your inbox the way we treat our clients’ brands.