Short answer

Use it as a fast assistant, not as the author. Let it draft from facts you give it, tighten titles and descriptions, turn customer questions into clear answers and sketch structured data. Then check everything it writes, because it will sometimes state things that aren't true.

Good jobs to give it

  • Titles and descriptions. Paste a page and ask for five title options under 60 characters that say what the page offers and where.
  • Answers to real questions. Give it the questions customers ask you, with your actual answers in rough form, and ask for a two-sentence version of each.
  • Plain-English rewrites. Ask it to rewrite a dense paragraph so a customer understands it, without changing the facts.
  • A first draft of structured data. Give it your business details and ask for LocalBusiness or Organization markup. Then test it in Google's Rich Results Test before using it, and make sure every detail in it is true.

Jobs not to give it

  • Writing dozens of near-identical pages for different towns or keywords. Google warns this kind of mass generation “may violate Google's spam policy on scaled content abuse”.
  • Inventing statistics, reviews or case studies. If it isn't true, don't publish it.
  • Telling you how your site is doing. That needs real data, from Search Console and testing tools.

Check before you publish

Google is clear that “generative AI outputs may contain inaccuracies” and asks site owners to fact-check everything, including titles, meta descriptions, structured data and image descriptions (Google Search Central). A wrong price or a made-up service in your structured data is worse than none at all.

The part it can't do

It doesn't know your customers, your prices or what makes you different. Those are what make a page worth ranking and worth quoting, so they have to come from you.

Sources