How to Get Your Business Recommended by ChatGPT, Perplexity, Gemini

Here is what the 2025–2026 data says actually gets a business named in AI answers, and how to check your own.

Does ChatGPT Actually Recommend Businesses?

Yes, but less often than most owners assume. Forrester's October 2025 Consumer Pulse Survey found that 19% of US online adults had used ChatGPT in the past month to search for a product, barely up from 18% in February 2025. Google (69%) and Amazon (58%) remain far more widely used for product discovery, and Perplexity actually slipped from 4% to 3% over the same period (Forrester).

That matters for how you plan. AI assistants are a real, growing channel, not yet a dominant one. Answer engine optimization (AEO) is the practice of shaping your content so assistants like ChatGPT can find, quote, and recommend it. Generative engine optimization (GEO) is the closely related term for the same work, usually used for AI search tools broadly, including Google's AI Overviews. Both terms describe the same job: making your business the easy, trustworthy answer when someone asks an AI tool a buying question.

How Does ChatGPT (and Perplexity, and Google) Decide What to Cite?

ChatGPT, Perplexity, and Google AI Overviews each pull from a mix of retrieved web pages and, in ChatGPT's case, some built-in training knowledge, then compose an answer with a handful of citations. Which pages make that cut changes over time and by platform, so a single "trick" rarely works for long.

Semrush tracked this shift directly. Its 13-week study covered more than 230,000 prompts and over 100 million citations across ChatGPT search, Google AI Mode and Perplexity (July 14 to October 12, 2025). In early August 2025, Reddit was cited in close to 60% of ChatGPT responses and Wikipedia in roughly 55%. By mid-September, Reddit's share had fallen to about 10% and Wikipedia's to under 20%, while PR Newswire, Forbes, and Medium gained ground (Semrush). In the same study, Perplexity's citation mix stayed more stable, led by Reddit, LinkedIn, NIH, Microsoft, and Google.

One popular fix does not reliably work: schema markup alone. Ahrefs tracked 1,885 pages that added JSON-LD schema against 4,000 matched pages that didn't, from August 2025 to March 2026. Result: no meaningful citation lift on ChatGPT, Google AI Mode, or AI Overviews (Ahrefs). Schema can help a search engine parse your page correctly, but every page in this sample was already heavily cited (100+ AI Overview citations), so the study cannot say what schema does for pages AI systems do not yet surface.

Getting cited inside the answer, not just ranked below it, is also becoming the whole game. Pew Research found that when a Google AI summary appears, people click a traditional search result in just 8% of visits, versus 15% when no AI summary appears, and only 1% of visits include a click on a link inside the summary itself (Pew Research Center). If your business isn't one of the sources named in that summary, you are left competing for the 8% of visitors who still click a regular result, while everyone who stops at the AI answer never sees you.

What does move the needle, per a Princeton-led Generative Engine Optimization study (with co-authors from IIT Delhi and independent researchers): adding clear statistics, direct quotes, or cited sources to a page. In the study's benchmark, the top-performing methods (cite sources, add quotations, add statistics) each improved visibility by roughly 30-40% on the paper's position-adjusted word count metric and 15-30% on its subjective impression metric. They were tested as separate methods, not one combined change, and the best method varied by topic (Aggarwal et al., arXiv). In plain terms, the 2023 benchmark rewarded pages with specific numbers, quotations and cited sources.

Where This Shows Up: Use Cases Across ChatGPT, Perplexity, and Gemini

AI recommendations show up at different points in a buyer's search, not just "best company near me." Common patterns for high-ticket service businesses:

• Comparison questions. "What's the difference between a fee-only financial advisor and a commission-based one, and who does that well in [city]?" This rewards businesses with clear, specific comparison content, not just a homepage.
• Problem-first questions. "My roof is leaking after a storm, who handles insurance claims for storm damage?" This favors businesses whose content answers the underlying problem, not just the service name.
• Vetting questions. "Is [company] legitimate, what do reviews say?" Reviews and directory listings are among the outside sources an AI answer can draw on for these questions.
• Local plus AI Overviews. A Google search that triggers an AI Overview can cite a local business directory, review aggregator or news mention instead of the business's own site.
• Perplexity-style research questions. "Compare the top three commercial insurance brokers for a 20-person manufacturing company." In Semrush's data, Perplexity's most-cited domains were Reddit, LinkedIn, NIH, Microsoft and Google.

Vetting questions are especially common in regulated, high-trust categories. We ran this exact pattern at scale for one niche: see who AI actually names as the best immigration consultant in Canada, and why the answer changes by engine and by the applicant's home country. If you run that kind of regulated practice yourself, our immigration consultant marketing playbook applies the same data to the operator side: what to fix first so AI assistants actually name your firm.

What Our Own Data Shows: One Niche, 99 AI Answers

Instead of a made-up case study, here is a real snapshot. On 24 September 2026 we ran 20 questions that applicants ask about Canadian immigration consultants through ChatGPT, Gemini, Perplexity, Copilot and Google AI Mode, inside the Glasa product. We got 99 answers; 78 could be fully analyzed. (Full research.) For the complete picture on this one niche, see our AI visibility guide for Canadian immigration consultants.

Four patterns stood out, and each one maps to a step in the plan below:

• Nobody owns the category. The answers named 200+ different firms, most of them only once. The most-named firm showed up in 24 of 78 answers (31%). In a fragmented market, a small firm can still win specific questions.
• The most-named firm is also a top source. Its own website was the second most-cited domain in the whole study (26 answers). Publishing guides that answer buyer questions feeds the answers that recommend you.
• Third-party sites carry a lot of weight. The most-cited domain was a comparison site, coursecompare.ca (28 answers), ahead of every firm's website. Directories, review sites and Reddit followed.
• Trust signals come first. 87 of 99 answers mentioned the official regulator, and most told the buyer to verify the consultant's licence. If your category has a register, licence or certification, make it easy to verify.

This is one run on one day, and AI answers change between runs, so treat it as a snapshot rather than a ranking. The patterns, though, are the ones the steps below are built around.

How to Get Your Business Recommended by ChatGPT: A Step-by-Step Plan

1. Run a baseline check. Ask ChatGPT, Perplexity, and a Google search that triggers an AI Overview the 10-15 real questions your buyers ask before hiring someone like you. Write down who gets named.
2. Fix the content gap first. Rewrite your core service pages around the buyer's actual question, with specific numbers, timelines, and pricing ranges. Vague marketing language gives an AI system little to quote.
3. Grow specific reviews. Ask recent clients for specific reviews that name the service they used.
4. Earn third-party mentions. Get listed in real industry directories, trade associations, and local press. In the Semrush data above, the most-cited domains were third-party sites such as Reddit, Wikipedia, LinkedIn, Forbes and PR Newswire.
5. Add structured data, but don't stop there. Implement schema markup for your services and FAQ content. Treat it as basic hygiene for Google, not a guaranteed AI-citation lever, given the Ahrefs schema test above.
6. Add statistics and cited sources to your own pages. Following the Princeton GEO findings, put real numbers and named sources directly on your site, not just in a PDF report visitors have to download.
7. Re-run the baseline check monthly. AI answers change as platforms update their models and sources. A one-time fix does not stay fixed.

How Much Does AI-Visibility Tracking Cost?

Dedicated AI-visibility tools range from about $29 to $489 per month (checked 28 October 2026); see Glasa pricing. Manual checking is free but does not scale past a handful of prompts, and it is easy to forget to re-run. Dedicated tools automate the prompt-running and track which sources get cited over time.

For ROI, weigh the tool cost against staff time spent manually re-typing the same 10-20 prompts every month across three or four AI tools. Glasa is a similar tool in this category: it tracks how AI platforms such as ChatGPT, Gemini and Perplexity answer questions about your market and brand, and shows brand mentions, competitors named alongside you and the sources cited.

When This Isn't a Good Fit

• You have no reviews or third-party mentions yet. AI systems lean on outside sources more than your own site. If you have almost no outside footprint, fix that before chasing AI citations specifically.
• Your buyers don't research this way. Some high-ticket categories still move mostly through referrals and cold outreach. Check your own analytics and client interviews before investing heavily; do not assume AI-assistant traffic based on industry-wide averages alone.
• You need guaranteed, fast results. The Semrush citation-mix data shows this shifts with platform changes you don't control, sometimes abruptly, not on a schedule you can guarantee.

Common Mistakes That Keep Businesses Invisible to AI Answers

1. Treating schema markup as the whole strategy. As the Ahrefs test above shows, schema alone did not produce a citation lift; it's a support tool, not a fix for thin content.
2. Writing for search engines, not for the actual question. AI answers favor pages that open with a direct answer to a specific question, not a general company overview.
3. Ignoring reviews and third-party listings. If nothing outside your own site mentions you, there's little for an AI system to cross-check or cite.
4. Checking once and stopping. Citation patterns shift with platform updates, as the Semrush before/after data shows. A single audit goes stale within months.
5. Copying competitor claims into your own answer without a source. Turning "studies show" or a competitor's number into your own copy, without checking or linking the original source, is a good way to publish something wrong.

Not legal or financial advice. Data from one run on 24 September 2026; AI answers change.

Sources

• Forrester: US Consumers' Adoption of Answer Engines for Product Discovery Remains Flat (December 2025, October 2025 survey)
• Pew Research Center: Google users are less likely to click on links when an AI summary appears (July 22, 2025)
• Aggarwal et al., "GEO: Generative Engine Optimization," arXiv (November 2023)
• Semrush: The Most-Cited Domains in AI, a 13-week study (November 10, 2025)
• Ahrefs: Does Schema Markup Predict AI Citation? (May 11, 2026)
• Otterly.ai pricing
• Profound pricing