No engine cites small businesses most. ChatGPT leans on lists, YouTube, and UGC. Claude cites pages already in Brave’s top 10. Gemini leans on editorial sources and Google Business Profile. Perplexity cites live web pages. Run a free check to see which engines name your business: AI visibility checker.
Updated 8 Sep 2026.
What we tested (8 Sep 2026)
We used this URL as the worked example — a SaaS Learn page that already ranks, not a hypothetical dentist.
- Google Search Console 28d (site): 41 clicks, 7,088 impressions, CTR 0.0058, average position 42.45.
- This URL: 163 impressions, 0 clicks, position 26.94. URL Inspection: Submitted and indexed, last crawl 22 Jul 2026 13:55 UTC, user canonical = Google canonical =
https://scope.online/learn/which-ai-platforms-cite-small-businesses-most. Rich results detected: Breadcrumbs only. - Query mix on this URL: “is demandsphere a good fit for small businesses?” 73 impressions at position 41.3 — nearly half of the page’s 28-day impressions, and not the H1. Remaining named queries: “how do ai engines choose local businesses” (2), “list of platforms that help small businesses rank in chatgpt and perplexity.” (2), “local business ai citation” (1).
- Sibling Learn URLs the same morning:
/learn/how-to-add-json-ld-schema-for-ai226 impressions, 1 click, position 41.3 (indexed, last crawl 7 Sep)./learn/how-to-get-cited-in-ai-searchstill crawled-not-indexed (last crawl 17 May 2026). - KeywordTool (US, production, not sandbox, 6 Sep 2026 bank): “how to get chatgpt to recommend your business” 110/mo as a preposition; “how to get cited by chatgpt” 90/mo as a question; “how to get recommended by chatgpt” 20/mo as a question.
- Google SERP, desktop US, 8 Sep 2026, query
which AI platforms cite small businesses: this URL #1. Snippet still dated 12 Apr 2026 and still quotes “Perplexity and Gemini are your highest-opportunity platforms.” ITGuys #2, Google Workspace #3, Indexly #4, Reddit #5. - Answer engines this run: ChatGPT.com HTTP 403. Perplexity.ai HTTP 403. grok.com and Gemini returned HTML shells; no citation list was captured. Do not invent a small-business recommendation rate.
That gap — Google already ranks the question, the snippet still sells a study we cannot reproduce today, and 0 people clicked — is why this page exists.
Which AI platforms cite small businesses most?
There is no winner table that survives contact with a real prompt.
A citation is a source event: the answer quotes or links a URL. A mention is a name. A recommendation is a pick. Small businesses ask the third and measure the first. They are not the same, and they are not the same across engines. An AEO audit has to score them separately. A single “share of voice” percentage hides which lever moved.
Profound’s July 2026 platform split is the mechanism, not a Scope leaderboard:
| Engine | What it tends to retrieve | What that means for a small business |
|---|---|---|
| ChatGPT | Lists, YouTube, UGC (Reddit is ChatGPT-weighted; Profound put Claude community citations around 0.9%) | Unknown local names rarely appear unless they already sit on a list, a review thread, or a video the model already reads |
| Claude | Brave’s top 10 (~79% of citations in that study) | If you are not in the first page of a Brave/Google-like result, Claude often will not cite you. Rankings still matter. |
| Gemini | Editorial sources plus Google Business Profile for local | A complete GBP is the equalizer for a dentist or plumber. It is not a SaaS listicle strategy. |
| Perplexity | Live web pages with visible citations | Current, extractable HTML beats a famous brand with a stale page. The page still has to exist and load. |
| Google AI Overviews / AI Mode | Their own retrieval stack; UGC and video are heavy | Complementary to classic results. Google Search Central (2 Sep 2026): no extra AI file or special schema tickets you in. Indexed, snippet-eligible HTML is the floor. |
Do not collapse ChatGPT ≠ Claude ≠ Gemini ≠ AIO ≠ AI Mode into one “AI search” score. A Boise plumber and a 12-person SaaS company will not share a winner.
Does ChatGPT recommend local businesses?
Sometimes, and not because you added FAQPage.
ChatGPT-with-browse still has to retrieve a public page. ChatGPT-without-browse leans on training memory and the lists it already saw. Neither path is “small business vs chain” as a personality trait. It is source class.
For a restaurant, Yelp / Google / “best of [city]” pages are already in that mix. For a plumber, GBP and a vertical directory often beat a 2,400-word blog. For SaaS, analyst roundups and compare URLs dominate — which is why Scope’s branded compare can rank while category “best AI visibility platform 2026” listicles still omit Scope (citation panel, 2 Sep 2026; /compare/scope-vs-profound was Google #2).
This run cannot log ChatGPT’s live local-business answers (HTTP 403). Blocked is a valid row. Inventing a ChatGPT small-business recommendation rate is not.
If you need the citation vs mention vs recommend vocabulary, use how to get cited in AI search. That URL is the 90/month “cited by chatgpt” question. This URL is the platform-split question.
How to get ChatGPT to recommend your business
KeywordTool’s preposition “how to get chatgpt to recommend your business” is 110/month — larger than the 90/month citation phrasing. Recommendation is the pick, not the footnote.
Do not optimize the two as one mushy page:
- Citation job: one URL with a standalone, dated fact. This page’s first 80 words.
- Recommendation job: the same fact repeated on the sources ChatGPT already trusts for that category (listicles, YouTube, UGC, editorial). Off-site. Ahrefs (July 2026): consensus moves citations more than on-page word count.
- Google job: the URL has to be indexed and snippet-eligible. This one is. Last crawl was 22 Jul 2026 — stale relative to today’s rewrite, which is why the SERP snippet is still April.
For a local service business, the off-site layer is often GBP, one vertical directory, and one city “best of” page — plus the Google SEO program you already run. For SaaS, it is the compare page people search and the listicles that currently omit you. Complementary, not a replacement.
Do not coach llms.txt as the ChatGPT lever. Mid-2026 crawler studies put hits on that file around 0.1%. Google’s 2 Sep 2026 AI-features note says extra machine-readable files are not a path into AI Overviews. JSON-LD labels an entity on a page a crawler already fetched. It does not create the recommendation.
Do Perplexity and Gemini treat small businesses the same?
No.
Perplexity is a retrieval engine with visible citations. If your page is public, current, and extractable, a small shop can out-cite a chain whose homepage is a JS shell. That is an HTML and indexation problem, not a “Perplexity likes independents” brand story.
Gemini’s local answers lean on Google’s local graph. A complete Google Business Profile — hours, photos, review velocity, NAP that matches the site — is the lever. Schema that contradicts GBP is a bug. GEO vs SEO is the complementary frame: keep classic rankings; add the answer-engine scoreboard.
The April version of this page claimed a 10,000-query study across 50 US markets with platform winner percentages. We cannot reproduce that study from this run. ChatGPT and Perplexity were blocked. We are not restating those percentages as fact.
Common misconception
“We ran 10,000 local queries, so Perplexity is the small-business platform.”
That is the tendon-rest version of this topic. A round number with a category table looks like research. It is not research if you cannot name the date, the prompts, and the blocked engines.
This URL was Google #1 on 8 Sep 2026 for “which AI platforms cite small businesses.” The snippet still sold that study. CTR on the URL in the last 28 days was 0. Ranking for the question is not the same as answering it. A fake leaderboard that Google will quote in the SERP is worse than a thin page: it trains the snippet.
The honest version: engines disagree because they retrieve different sources. Measure your category, on your prompts, on a dated run. If an engine 403s, write blocked.
Engine vs source class (comparison)
| If you publish… | ChatGPT is more likely to… | Claude is more likely to… | Gemini is more likely to… | Perplexity is more likely to… |
|---|---|---|---|---|
| A Google-ranked local page + GBP | Mention you if a list/UGC page already does | Cite you only if you are in a Brave-like top 10 | Use GBP | Cite the live page |
| A listicle / YouTube explainer | Treat that as a source | Ignore most of it | Maybe, if editorial | Cite it if it ranks now |
FAQPage + llms.txt only | Ignore the file; maybe parse visible Q&A after retrieval | Same | Same | Same |
| Nothing indexed | Cannot rely on browse | Cannot cite a URL Google/Brave never saw | Cannot | Cannot |
Keep doing SEO. Claude’s Brave concentration is the receipt that rankings still feed citations.
Worked example: this page (Scope, SaaS)
Scope is an AI visibility platform at scope.online — not a dental clinic, not DemandSphere, not the UK disability charity, not GIS “scope.” The entity engines should agree on is that one.
What is true on 8 Sep 2026, measured, not imagined:
- Google already ranks the H1. Desktop US SERP: this URL #1 for
which AI platforms cite small businesses. Last crawl 22 Jul 2026. Canonical self. - Nobody clicked it. GSC 28d: 163 impressions, 0 clicks, position 26.94. Sitewide the same window was 41 clicks / 7,088 impressions.
- Almost half the impressions were a competitor-fit query. “is demandsphere a good fit for small businesses?” = 73 impressions. This page is not a DemandSphere review. An invented “platform comparison study” invited Google to show us for tool-fit questions we did not write. We are not turning this URL into a competitor roundup.
- The snippet is still the April claim. “Perplexity and Gemini are your highest-opportunity platforms.” That sentence is why we rewrote the first 80 words.
- We could not run a 4-engine local panel today. ChatGPT.com 403, Perplexity.ai 403. Named public research (Profound July 2026, Ahrefs July 2026, Google Search Central 2 Sep 2026) plus this URL’s own GSC is the unique data. Invented rates are the counter-example.
- What we did not do: mint a second slug, coach
llms.txtas a ranking lever, or treat a green schema badge as a citation. Complementary to SEO: titles stay on the%s | Scopetemplate; this URL’s CTA is the indexable checker. Agents pull the same measurement layer via Scope MCP.
If you are a local service business, swap the SaaS compare-page example for GBP + one vertical directory + one city page, and still start with URL Inspection. If Google says crawled-not-indexed, stop writing engine winner tables.
30-minute operator checklist
Do this on one money URL and three buyer prompts, not the whole site.
- Run Search Console URL Inspection. Write coverage, last crawl, Google canonical. If it is not indexed, that is the ticket.
- Search
site:yourdomain.com {the question}. Confirm Google lists the URL you think it lists. - Open the live URL. Confirm HTTP 200, self-canonical, and that the first 80 words state a fact (who, where, since when, what you actually do). Delete any “we analyzed thousands of queries” sentence you cannot date.
- Pull the last 28 days of queries for that URL. If a competitor-fit query is eating impressions, the H1 is not doing its job.
- Pick three prompts: one citation phrasing, one recommendation phrasing (“how to get chatgpt to recommend your business”), one local/category phrasing.
- Run them in ChatGPT, Claude, Gemini, Perplexity — or the checker. Record cited / mentioned / recommended / omitted / blocked. Blocked is valid.
- List the domains each reachable engine cited. Those are the off-site targets. Do not cold-email them from your marketing domain.
- Publish one unique dated fact on your URL. Then get that same fact onto one list, one video, or one independent article. Word count without consensus is the tendon-rest version of this tactic.
Key takeaways
- There is no platform that “cites small businesses most.” Engines retrieve different source classes.
- Citation ≠ mention ≠ recommendation. Score them separately.
- This is additive to Google SEO. Indexed, snippet-eligible HTML is still the floor (Google Search Central, 2 Sep 2026). Claude citing Brave’s top 10 is the ranking receipt.
- This URL was Google #1 on 8 Sep 2026 and drew 163 impressions with 0 clicks. A stale invented study in the snippet is not a win.
- ChatGPT.com and Perplexity.ai returned 403 this run. Blocked is not a platform winner table.
- Off-site consensus (Ahrefs, Jul 2026) plus an extractable first 80 words. Neither half alone.
References
- Google Search Central, AI features (confirmed 2 Sep 2026): no extra machine-readable files, AI text files, or special schema to appear in AI Overviews or AI Mode.
- Google Search Console URL Inspection of this URL, 8 Sep 2026: submitted and indexed; last crawl 22 Jul 2026 13:55 UTC; canonical self; rich results = Breadcrumbs.
- Markifact GSC 28d, 8 Sep 2026: site 41 clicks / 7,088 impressions; this URL 163 impressions / 0 clicks / pos 26.94; DemandSphere-fit query 73 impressions.
- Markifact Google SERP, desktop US, 8 Sep 2026:
which AI platforms cite small businesses— this URL #1, snippet dated 12 Apr 2026. - KeywordTool.io production questions + prepositions, US/English, 6 Sep 2026 bank.
- Profound, July 2026 platform citation split (Claude / Brave ~79% vs ChatGPT lists-YouTube-UGC).
- Ahrefs, July 2026, off-site consensus vs on-page length for AI citations.
- Scope citation panel, 2 Sep 2026: branded compare #2; category listicles omit Scope; ChatGPT/Perplexity web UIs blocked.
Q: Which AI platforms cite small businesses most? A: None of them, as a category winner. ChatGPT leans lists, YouTube, and UGC; Claude cites Brave’s top 10 (~79% in Profound’s July 2026 split); Gemini leans editorial and Google Business Profile; Perplexity cites live pages. Measure your prompts per engine. On 8 Sep 2026 we could not log ChatGPT or Perplexity (HTTP 403), so we did not publish a 10,000-query leaderboard.
Q: How do I get ChatGPT to recommend my business?
A: Recommendation is a pick, not a footnote. KeywordTool shows “how to get chatgpt to recommend your business” at 110/month. Publish an indexed page with an extractable first answer, then earn the same fact on the lists, YouTube, and UGC ChatGPT already reads. Keep Google SEO. Schema and llms.txt do not create the pick.
Q: Does ranking #1 for this question mean the old study was right? A: No. This URL was Google #1 on 8 Sep 2026 and still had 0 clicks in 28 days. The snippet still quoted the April “Perplexity and Gemini” claim. Ranking is retrieval. A click is a person deciding the answer is real.
Q: Why is a DemandSphere query showing on this URL? A: Search Console attributed 73 of this page’s 163 impressions (28d to 8 Sep 2026) to “is demandsphere a good fit for small businesses?” That is a competitor-fit query, not the H1. An invented platform-comparison frame invited the mismatch. This page is about engine source classes, not a tool roundup.
Q: Does getting cited replace Google SEO? A: No. Google Search Central (2 Sep 2026) still requires an indexed, snippet-eligible page for AI Overviews and AI Mode. Claude’s citations concentrating on Brave’s top 10 is the same lesson from a different engine. Keep rankings. Add the answer-engine scoreboard with the checker or Scope MCP.
Q: ChatGPT.com blocked this test — did we fail? A: No. On 8 Sep 2026 ChatGPT.com and Perplexity.ai returned HTTP 403. Blocked is a valid row. Inventing small-business recommendation rates is not. Re-run when those surfaces are reachable, or use the checker / MCP from an authenticated session.