How to Get AI to Say Good Things About Your Business (Without Gaming It)

To get AI to say good things about your business, change what its sources say: fix the recurring complaint behind any negative point, reply to reviews (critical ones first), collect a steady flow of recent genuine reviews, and give the web specific proof of what you do well, in your own pages and in coverage by others. Assistants summarise the reviews, threads, articles and pages they read. They don’t form an opinion of their own, so there is no setting to change, and attempts to fake the inputs tend to be caught.

Where the tone of an AI answer comes from

Gemini in Google Maps makes the mechanism explicit: Ask Maps builds its recommendations from more than 300 million places and reviews from more than 500 million contributors (AP, March 2026). What reviewers say about your place is the material it works with.

Step 1: find the theme behind the negative answer

1. Ask the assistants “Is [business] any good?”, “What do reviews say about [business]?” and “[business] vs [competitor]”, three times each, with search on.
2. Write down every negative point and the source it cites.
3. Read your last 50 reviews and tag the complaints. In most businesses, two or three themes account for most of the negative ones: wait times, communication, a pricing surprise, one product.

If the negative point traces to something that’s no longer true (an old owner, a fixed problem, a closed location), it’s a wrong-information problem; follow how to fix wrong AI information. If it’s still true, the fix is in the business, not in the marketing.

Step 2: fix it, then say so

Fixing the complaint changes what the next reviews say, and recency matters: in BrightLocal’s 2026 survey, 74% of consumers look only at reviews from the last three months, and Ahrefs found AI assistants cite newer content than Google’s organic results do (vendor study). Then make the fix visible: mention it in replies to the reviews that raised it, add it to the relevant page (“same-day callbacks since March”), and tell the customers who complained, some of whom will update their reviews.

Step 3: reply to reviews like someone will quote you

They might. A reply sits next to the review it answers and is read with it, by people and by the systems that summarise. 89% of consumers expect owners to respond (BrightLocal). Good replies are short, specific and calm: thank them, acknowledge the actual problem, say what you changed or how to reach you, and stop. Never argue, never reveal a customer’s private details, never offer a reward for changing a review.

Step 4: publish proof, not praise

Assistants drop adjectives and keep facts. “Award-winning, best-in-class service” gives them nothing; “4.8 stars from 1,200 Google reviews, 94% of jobs finished on the first visit, Gas Safe registered since 2011” gives them a sentence to repeat. Self-praise can even work against you: Lily Ray’s analysis of Ahrefs Brand Radar data covered 100 B2B queries: 69% of the self-promotional “best” lists Google’s AI Overviews cited (224 of 323) belonged to brands the answer didn’t recommend. In the AutoGEO research (ICLR 2026), simulated AI engines preferred specific, neutral, current content; those were small test engines, not production assistants, so read it as a direction rather than a measured effect.

• Case studies with numbers, named clients (with permission) and dates.
• Credentials, registrations and memberships, with links to the issuing body.
• Awards and coverage from independent sources, linked.
• Published prices and policies, so assistants don’t fill the gap with complaints about surprises.

Step 5: get good things said by others

Praise on your own site is a claim; praise elsewhere is evidence. Ask every satisfied customer for a review (how to, within the rules), pitch the lists that cover your category (best-of list outreach), and take part honestly where your customers talk (Reddit and AI search). Google’s guide is direct about shortcuts: “seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem”.

What never to do

• Fake reviews, bought reviews or staff posing as customers. The FTC’s 2024 rule bans them, including AI-generated ones, and allows civil penalties.
• Suppressing genuine negative reviews through threats or intimidation, which the same rule bans.
• Asking only happy customers for reviews. Google’s policy prohibits selectively soliciting positive reviews.
• Planting “independent” comparison pages or forum posts. Moderators and platforms look for exactly that, and being caught is the kind of story assistants repeat for years.

Measure the change

Keep the same three questions and re-ask them monthly. Track two things: how many negative points each answer contains, and whether the sources behind them are older or newer than your fix. The question set and scoring sheet are in what AI says about your business.