The same conversation happens almost every week now. A firm owner opens ChatGPT, types something like "best estate planning attorney in Scottsdale," watches it confidently name three competitors, and calls me wanting to know what those firms did to get in there. Usually the answer is nothing deliberate. They just happened to look, to a machine reading the public internet, like the obvious answer.
That's the part worth sitting with. AI assistants don't have a sales team you can call and they don't take submissions. They assemble recommendations from what's already written about you, which means getting recommended is less like buying an ad and more like building a reputation the machines can read.
What the model is doing when someone asks for a recommendation
When someone asks ChatGPT for a lawyer or an advisor, the model isn't querying some official directory. Depending on the product and the question, it's doing a blend of two things. It recalls patterns from its training data, years of the public web where certain firms kept showing up in reviews, directories, local press, and everything else it read. And increasingly it runs a live web search behind the scenes and synthesizes the top results into one confident answer. (The mechanics of how those answers get built, and what they mean for your existing SEO, are their own topic. I covered that in /blog/ai-seo-vs-traditional-seo/.)
Both paths reward the same underlying condition. If your firm is described consistently, reviewed heavily, mentioned by trusted sources, and easy to crawl, you show up. If your web presence is thin, or the important parts are locked inside widgets a crawler can't parse, you don't exist as far as the answer is concerned.
And the stakes aren't theoretical. Semrush measured AI search visitors converting at roughly 4.4 times the rate of traditional organic visitors, and about 6 in 10 consumers now say AI answers influence what they buy. The person who asks ChatGPT for an attorney and gets your name is far closer to hiring you than the person who clicked your link on page one of Google.
The order I'd fix things in
I've watched enough firms work through this to have a strong opinion about sequence. Don't start with the clever stuff. Start with the boring stuff, because the boring stuff is what the models actually weight.
1. Reviews, before anything else
Reviews are the closest thing to a training signal you can directly influence. Models learn who's good the same lazy way humans do: volume, rating, recency, and the actual words people use. A firm with 240 Google reviews that keep mentioning "estate planning" is teaching every future model what it does and how well it does it. A firm with 11 reviews from 2021 is teaching it nothing.
If you're wondering what the target number even is, I did that math separately in /blog/how-many-google-reviews/. Short version: more than you think, arriving steadily rather than in bursts.
2. Make your business information boringly consistent
Same name, same address, same phone number, same service descriptions, everywhere. Your Google Business Profile, your site footer, your state bar or FINRA listing, whatever directories matter in your vertical. Inconsistency doesn't just cost you a little credit. It can split you into two fuzzy entities in the model's understanding, neither strong enough to recommend. This is old-school local SEO hygiene, and it matters more now, not less.
3. Add schema markup so machines don't have to guess
Structured data (LocalBusiness or LegalService markup, plus FAQ markup on pages that answer real client questions) is you spelling out, in a format built for machines, exactly what you do and where you do it. It takes a developer about an afternoon. I'm consistently surprised how many firms doing seven figures a year have skipped it.
4. Get mentioned in places AI reads
Quotes in local news, chamber of commerce pages, industry roundups, podcast show notes, "best in the city" lists that aren't pay-to-play. One quote in a regional business journal does more for your AI visibility than ten blog posts nobody links to, because third-party mentions are how a model separates firms that say they're good from firms other people say are good. This is the slowest lever to pull, which is exactly why the firms that started early are so hard to displace.
5. Make sure a crawler can actually read your site
Plain HTML with real text on the page, and service pages that name your city and practice areas the way a client would say them. Nothing load-bearing trapped inside images or scripts. While you're at it, check that your robots.txt isn't blocking GPTBot or the other AI crawlers, a leftover developer default I find on maybe one site in five.
What doesn't work, so you can stop paying for it
Keyword stuffing does nothing here. The models aren't matching strings, they're summarizing consensus, and a page that repeats "best divorce lawyer Houston" fourteen times reads as spam to a human and to a model trained on humans.
Begging doesn't work either. You can't email OpenAI and ask to be included, and there's no submission form because there's no index to submit to, just the accumulated public record of your firm. Anyone selling "guaranteed ChatGPT placement" is selling weather.
Fake reviews are worse than useless. Google's filters catch most of them, the FTC now fines for them, and a suspiciously perfect rating pattern is exactly the kind of anomaly a model can learn to discount.
The honest timeline
None of this is instant, and I'd rather you hear that from me than from your invoice six months in. Review velocity compounds over months and authority mentions take quarters. Training data refreshes on the model provider's schedule rather than yours, though the live-search paths pick up changes much faster. The firms getting named today mostly did the work in a previous year, which is the strongest argument I know for starting in this one.
If you want to know exactly where you stand before touching anything, we run a free AI Visibility Audit at https://scalewithquail.com/ai-visibility/ that shows what ChatGPT, Gemini, Perplexity, and Google's AI Overviews currently say when someone asks for your kind of firm in your city. It's a strange feeling, reading it the first time. Better to have that feeling now, while the answer is still cheap to change.
Questions people ask
How does ChatGPT decide which businesses to recommend?
It draws on patterns from its training data plus live web searches, and both reward the same things. Review volume and ratings, consistent business information across the web, mentions from sources it treats as authoritative, and a site its crawlers can actually read.
Can I pay to be recommended by ChatGPT?
No. There's no ad product or submission form for organic AI recommendations, and anyone selling guaranteed placement is guessing. What you can influence are the public signals the models learn from, starting with Google reviews and consistent business data.
How long does it take to show up in AI recommendations?
Answers that use live search can reflect changes within weeks, but recommendations grounded in training data move on the model provider's schedule, often months. Most firms see real movement after one or two quarters of consistent review and citation work.