The question that used to go to Google
When a homeowner in Dallas needs a roofer after a hailstorm, they used to type "best roofer Dallas" into Google and pick from the map pack. In 2026, more of them open ChatGPT instead and ask "who should I call for roof damage after the hail last week". ChatGPT responds with a specific business name, a short case for why, and sometimes a link. The buyer picks from a shortlist of one or two. If your business is not on that list, you lost the deal before you knew the deal existed.
Why ChatGPT picks specific businesses
ChatGPT does not pick randomly. Behind every named recommendation is a weighted combination of signals the model learned during training and reinforces through live browsing when enabled. The six signals that matter most, in rough order of weight: (1) structured entity presence (schema.org, Wikipedia, Wikidata), (2) review consensus across sources (not just Google — BBB, Yelp, Angi, forums), (3) citation density on sites ChatGPT crawls heavily (Reddit, Quora, trade press), (4) source diversity (one strong citation beats ten weak ones), (5) factual consistency across those citations, (6) freshness.
Signal 1: structured entity presence
ChatGPT disambiguates businesses the same way Google does — by comparing structured facts across sources. A business with LocalBusiness schema deployed on its website, a complete Google Business Profile, a Wikidata entry, and consistent name/address/phone across directories gets understood as one authoritative entity. A business with inconsistent NAP and no schema reads as noise.
Signal 2: review consensus
Review volume on Google alone is not enough. ChatGPT weights consensus — the agreement between reviews across multiple independent sources. Three hundred reviews on Google plus zero elsewhere reads as suspect. Eighty on Google, forty on BBB, twenty-five on Angi, and some positive Reddit mentions read as real. Build review surface area, not just review count.
Signal 3: citation density on trusted sources
ChatGPT reads Reddit, Quora, trade press, and local news during training and (when enabled) during browsing. Being mentioned, discussed, or recommended on those sources contributes directly to being named. This is where seeding — honest, high-signal participation in relevant communities — comes in. Not spam. Substantive help that happens to put your name in the conversation.
Signal 4: source diversity
A business cited by GAF (manufacturer), by a local news story, by a Reddit thread, and by BBB reads as more legitimate than a business cited only by its own blog and twelve directory listings. Diversity signals that external parties — not just the business itself — vouch for it.
Signal 5: factual consistency
If your Google Business Profile says you serve Dallas and your website says you serve Dallas–Fort Worth and your Yelp page lists only Richardson, ChatGPT does not know which to trust. It discounts your business accordingly. Align your facts across every public surface before investing in content.
Signal 6: freshness
Dead-looking businesses get deprioritized. A GBP with no posts in six months, a website blog with nothing new in a year, and no social activity signals abandonment. Weekly GBP posts, monthly blog content, and quarterly refreshed photos signal life. Not every business needs to be posting daily, but something consistent beats nothing.
The 90-day plan
Days 1-14: audit all six signals, fix fact inconsistencies, deploy schema and llms.txt, complete GBP. Days 15-45: build review surface area across BBB, Angi, Yelp, and category-specific directories. Launch review flywheel (QR codes, follow-up templates). Begin honest community participation on Reddit and Quora. Days 46-75: content push on your website — two long-form posts a month, FAQ structured for AI extraction, HowTo content for procedural topics. Days 76-90: measure. Weekly probes of ChatGPT, Claude, Perplexity, Google AI, Gemini, and Copilot for your category and geo keywords. Compare to baseline. If the lift is clear, scale what is working.
What a successful outcome looks like at day 90
ChatGPT names your business for the primary category and city query. Perplexity cites your website as a source for 2-3 educational queries in your vertical. Google AI Overview pulls your FAQ or HowTo content into the answer block for related searches. Branded search volume (searches for your business name on Google) rises 20-40%. All of it is measurable and tracked in a dashboard.