You get AI to recommend your business by fixing four signals in a deliberate order over 30 days: make your business facts identical everywhere, publish a quotable offer with real numbers, build third-party proof on the sources the models already read, and rewrite your key pages so they answer the questions customers actually ask. Measure a baseline before you touch anything and re-measure at the end, or you’ll never know which move worked.
This is the third post in that thread. Why ChatGPT doesn’t recommend your business explains the gap between being known and being named; how to check if AI recommends your business confirms that gap in one evening. Below is what to do about it, with nothing that requires a rebuild or an ad budget.
How do I get AI to recommend my business?
Work through four signals in a fixed order over four weeks: consistent business facts first, then a quotable offer with real prices, then third-party proof, then question-shaped content. Start and end the month with the same measurement. The order matters more than the speed, because content changes land harder once the underlying data is clean.
Most owners do this backwards: they rewrite the homepage first, because that’s the visible part, and leave the messy directory listings and the missing prices for later. But a model that finds three different phone numbers for you has a reason to hesitate no matter how well the homepage reads.
One caveat before the plan. Nobody can guarantee you a spot in an AI answer, and any agency that promises one is selling something. You’re shifting a probability, not flipping a switch.
Which four signals do AI tools actually read?
Four things decide whether a model can build a recommendation around you: a concrete offer it can quote, third-party proof it can verify, business facts that agree across sources, and content shaped like real customer questions. Miss one and the model reaches for a competitor it can describe without risk.
| Signal | What the model is checking | What it looks like when it works |
|---|---|---|
| Quotable offer | Can it state what you do and what it costs? | Prices or ranges in on-page text, services described with nouns, not adjectives |
| Third-party proof | Does anyone besides you vouch for you? | Recent reviews with owner replies, mentions on Reddit, review platforms, trade press |
| Consistent business facts | Do your name, address, phone and numbers match everywhere? | Identical data on your site, Google Business Profile, Bing Places, Apple Maps, industry directories |
| Question-shaped content | Does a page answer what customers actually ask? | Question headings, a direct answer in the first sentences, then the detail |
The fourth signal has a measurable shape. SE Ranking’s analysis of 129,000 domains and 216,524 pages across 20 niches found pages with an average section length of 120–180 words between headings earned about 4.6 citations, against 2.7 for pages with sections under 50 words. Content updated within the previous three months averaged 6.0 citations versus 3.6 for stale pages. Short answer blocks under clear headings, kept current: that’s the format models lift easily.
The first signal is subtler than it looks. In our own pilot audit, a dental practice had its implant price published in plain text, and AI still didn’t cite it when asked what implants cost in that city. The number existed; the citable form — under a question heading, attached to the named service — did not.
Why does third-party proof carry so much weight in English-language answers?
Because English-language models lean hard on a small set of public, high-volume sources — Reddit, Wikipedia, review platforms, press wires — and those are where your reputation is either visible or absent. What other people say about you is the part of the story a model can verify, which makes it the part it trusts.
The effect is documented. SE Ranking’s study found domains with millions of brand mentions on Reddit averaged 7 ChatGPT citations, against 1.8 for domains with minimal Reddit presence. That’s a correlation, not proof that posting on Reddit causes citations, and the volumes at the top end belong to very large brands — but the direction is consistent.
Which sources matter depends on what you sell:
- B2B software and services. G2, Capterra, TrustRadius. G2’s April 2026 “Answer Economy” report, from a March 2026 survey of 1,076 B2B software buyers, found 51% now prefer to start research in an AI chatbot rather than Google, up from 29% eleven months earlier — and 45% named citations from software review sites as the signal that gave them the most confidence in an AI recommendation.
- Consumer and local businesses. Google reviews, Trustpilot, Yelp, and the place databases behind the assistants. SOCi’s 2026 Local Visibility Index, covering nearly 350,000 locations across 2,751 multi-location brands, found ChatGPT recommended just 1.2% of locations when asked for a local option, against a 35.9% appearance rate in Google’s local 3-pack. Those it did recommend averaged 4.3 stars — an association rather than a threshold to hit, but a telling one.
- Everyone. Trade press, podcasts and genuine industry roundups. Semrush’s three-month citation study, published in November 2025 across 230,000+ prompts and more than 100 million citations, found PR Newswire, Forbes and Medium were the biggest gainers on ChatGPT after a mid-September shift — the same shift that cut Reddit’s and Wikipedia’s share of ChatGPT responses from roughly 55–60% to under 20%.
That last number is the warning worth keeping. Citation sources move, sometimes sharply, and differently on each platform. Earn proof in several places your customers would find credible anyway, and you’re insulated when a model re-weights its diet.
The 30-day plan, week by week
The sequence is deliberate: measurement, then data hygiene, then offer, then proof and content. Reviews and mentions start early because they’re the slowest to accumulate.
- Days 1–3: measure your starting point. Ask ChatGPT, Gemini, Perplexity and Claude the same ten questions your customers ask, once with your business name and once without, in clean incognito sessions. Screenshot everything. How to check if AI recommends your business has the prompt set; how to measure AI visibility has the protocol for running it three times so you get a rate, not an anecdote. Skip this and the other 27 days are unfalsifiable.
- Days 4–10: unify your business facts. Name, address, phone, hours, service area and services should be identical on your site, Google Business Profile, Bing Places, Apple Business Connect, your social profiles and your structured data. You’ll find contradictions; almost everyone does. SOCi’s index found business profile information was only about 68% accurate on ChatGPT and Perplexity, against 100% on Gemini — a gap that starts with source data, not the model.
- Days 8–15: publish a quotable offer. Specific services, specific prices or ranges, specific turnaround times, in on-page text. Each service gets its own page or section with a heading, and the number appears in the first sentences under it — not in a downloadable PDF, not baked into an image. If a full price list is impossible, publish ranges or a “from” price.
- Days 10–30: start the proof engine. Set up a one-click review link and ask after every job done well; reply to every review, including the unflattering ones. In parallel, pick the two off-site places that matter for your category — a subreddit, a review platform, a trade publication — and start showing up there with something useful to say. This is the signal a competitor can’t replicate over a weekend, which is why it starts on day 10 rather than day 25.
- Days 20–30: rewrite key pages and clear the technical path. Question headings, a direct answer in the first two or three sentences, then the expansion. Confirm
robots.txtisn’t blocking GPTBot, ClaudeBot or PerplexityBot, and check that important content renders as text rather than script. How to optimize your website for AI search is the detailed checklist. Then re-run the exact measurement from step one.
This asks for about an hour a day and the willingness to publish your prices. It doesn’t ask for a redesign. If you’re a single-location or service-area business, AI visibility for small and local businesses covers the local specifics.
What will set you back
Three shortcuts look efficient and cost more than they save.
Fake reviews are now a federal matter in the US. The FTC’s Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024. It prohibits creating, buying, selling or disseminating reviews from people who don’t exist or never used the product — the FTC has said explicitly that AI-generated reviews are covered — plus undisclosed insider reviews and the suppression of negative ones. Knowing violations can draw civil penalties of more than $50,000 per violation, adjusted annually for inflation. Models also cross-check sources, so a sudden burst of five-star enthusiasm with nothing behind it reads as noise, not proof.
Astroturfing on Reddit gets your account and your brand burned. Reddit’s Content Policy requires authentic participation in communities you actually have an interest in and prohibits content manipulation. Drop-in promotional links, sockpuppet accounts and vote manipulation mean removals and bans, and moderators are good at spotting them. The version that works is slow: answer questions in your field for weeks, disclose who you are, and mention your product only when it genuinely answers what someone asked.
Keyword stuffing does nothing here. Repeating “best emergency plumber Chicago fast affordable” doesn’t persuade a model that reads for meaning, and it makes the passage harder to quote than a sentence that simply answers the question. You lose twice: with the model and with the human who closes the tab.
The subtler mistake is changing ten things at once. If visibility improves you won’t know which change did it, and you’ll keep paying for the nine that didn’t.
How do you know if the 30 days worked?
Re-run the identical prompt set, in the same conditions, three times, and compare your presence rate — how often you were named, as a percentage of runs — against your day-one baseline. One appearance proves nothing; models answer the same question differently each time. A rate moving from 20% to 50% is a defensible result.
Keep the dated screenshots from both ends of the month; you can’t reproduce today’s exact answer in four weeks, so the image is your evidence. How to measure AI visibility covers presence rate, share of voice and cited sources in full, including how to tell a real improvement from normal model wobble. Be patient with the trust layer: technical fixes can register as soon as the tools re-crawl you, while reviews and mentions accumulate over months. If the rate hasn’t moved by day 30 but you’ve done the work, that’s information too — it usually means the constraint is off-site proof rather than anything on your website.
If you’d rather not run the measurement and the diagnosis by hand, that’s the job our AI visibility audit does: the same questions across ChatGPT, Gemini, Perplexity and Claude, 39 criteria scored across six areas, and a prioritized fix list with a 30-day plan built for your business rather than a generic one. It starts at PLN 499 net (around $125), and if we don’t find at least five things worth fixing, you get a full refund. Not sure you need it yet? A free 20-minute consultation will tell you honestly, or start wider with our free 3-minute marketing audit.
FAQ
How long before AI actually starts recommending my business? It depends on where you’re starting and how quickly the tools re-crawl you, so no honest guide gives you a date. Consistency fixes and technical access can register within weeks. Reviews and third-party mentions build over months. That’s why the plan is 30 days of work with a measurement at each end, rather than a promise of results by a certain morning.
Do I have to publish prices to get recommended? You don’t have to, but businesses that do have an advantage. When someone asks a model “how much does X cost and who does it,” it names the businesses it can attach a number to. If a full price list doesn’t fit your work, publish ranges or a starting price with what’s included. That’s usually enough to become quotable.
Should I be posting on Reddit for my business? Only if you can do it as a genuine participant, and only where your customers actually are. The correlation between Reddit presence and AI citations is well documented, but the version that earns it is answering questions in your field over weeks, with your affiliation disclosed. Drive-by promotion violates Reddit’s Content Policy and gets removed, leaving you a burned account and no signal.
Can I do all of this without an agency? Yes. Every step above is doable in-house: the measurement, the data cleanup, the price page, the review routine, the rewrites. An audit earns its cost when you want to skip guessing which fix matters most, or when you’ve done the 30 days honestly and the number hasn’t moved. To weigh that up first, is an AI visibility audit worth it makes the case both ways.