Strategy 9 min read
How to Get Your Software (SaaS) Recommended by ChatGPT and Gemini
Software is the category where your own website matters most to AI answers. In one 2026 index, ChatGPT cited the recommended product’s own site 53% of the time. The rest of the answer is decided by a few publishers and comparison pages.
Founder, SEO·AEO·GEO
Builds SEO·AEO·GEO and writes the crawler, schema, and citation research published here.
Key takeaways
- ChatGPT and Gemini named a different top pick in one software category in three, in a 126-category index from August 2026. Track each engine separately.
- Your own site carries real weight for software: ChatGPT cited the recommended product’s own website 53% of the time, Gemini 13%.
- Review sites were a small share of citations (G2, Capterra and TrustRadius together about 5%); tech publishers, Reddit and integration directories were bigger.
- Publish pages for the questions buyers ask: use cases, comparisons, alternatives, pricing and integrations, with the facts in plain HTML.
- Answers differ by country: the top pick flipped between US and UK phrasing in about half of measurable categories.
To get your software recommended by ChatGPT and Gemini, publish pages that answer the questions buyers ask assistants (use cases, comparisons, alternatives, pricing, integrations) with the facts in plain HTML; get named on the comparison articles and “best X software” lists the assistants cite for your category; keep your description and pricing consistent everywhere; and track ChatGPT and Gemini separately, because they often disagree. Software is the one category where your own site carries real weight in AI answers, so start there.
Those figures come from GetIntel’s AI Software Index, reported by The Next Web in August 2026: about 80 buyer-phrased questions in each of 126 categories, 9,978 answers from the live ChatGPT and Gemini apps, collected on one day. GetIntel sells AI visibility software, and the data has no error bars, so treat it as a snapshot. It is still the largest public look at software recommendations specifically.
Which sources decide software answers
The most-cited domains in the index were techradar.com (1,114 citations), Reddit (807) and Zapier (418), with Gartner at 238. Review platforms were a small share. Ahrefs’ study of recommendation prompts found the same shape from another angle: “best X” lists were the largest group of cited pages, and for software, a brand’s own blog list appeared in 34% of ChatGPT answers where that brand was recommended. Landing pages were the most common first-party page type cited for software, at 37.2%.
| Source | Why it’s cited | What to do |
|---|---|---|
| Your own site | It states what the product does, for whom and at what price | Use-case, comparison, pricing and integration pages, readable without JavaScript |
| Tech publishers and comparison articles | They answer “best X for Y” directly | Pitch the authors with a checkable case; ask for corrections when they’re wrong |
| Reddit and community forums | First-hand opinions and complaints | Take part honestly and disclose (how) |
| Integration directories (Zapier and app marketplaces) | They describe what connects to what | Keep listings complete and current |
| Review platforms (G2, Capterra, TrustRadius) | Ratings and buyer quotes | Keep them current; useful to buyers and list authors |
| YouTube | Demos and comparisons | Long-form walkthroughs (YouTube and AI) |
The pages every software site needs
- Use-case pages, one per buyer job: “[product] for [role/industry]”, saying what it does for them, with a number or example.
- Comparison pages: “[you] vs [competitor]”, fair and current, saying who each suits.
- Alternatives pages: “[competitor] alternatives”, where you’re one honest option among several.
- A pricing page with real numbers. Assistants asked about cost will otherwise quote a review site or a forum guess.
- Integration pages: one per major integration, saying what syncs and how.
- A plain product description on the homepage: what it is, for whom, and from what price, in the first paragraph.
Keep all of it in server-rendered HTML. Many software sites are single-page apps, and Vercel found that none of the major AI crawlers render JavaScript, so text that appears only after scripts run may be invisible to them. Our [free audit](/) shows what each AI crawler receives from your pages.
Writing comparisons that help rather than hurt
Self-promotional lists split the evidence. Ahrefs found no clear downside for software brands that publish their own; in Lily Ray’s test of 100 B2B queries, 69% of the self-promotional lists AI Overviews cited (224 of 323) belonged to brands the answer didn’t recommend. The pages that work for buyers, and so for assistants, share three traits: they compare on facts a buyer can check (price, limits, integrations), they say plainly when a competitor is the better choice, and they carry a visible updated date that’s true.
Track ChatGPT and Gemini separately, by country
With different top picks in a third of categories and a US–UK flip in about half, one combined “AI visibility” number hides most of what matters. Set up your buyers’ questions per engine and per market, run them monthly, and record the top pick and the cited sources. Our AI visibility checker, rolling out engine by engine, shows who each engine recommends instead of you; the dashboard tracks the same questions weekly.
What about a ChatGPT app?
OpenAI opened submissions to its app directory in December 2025, and its docs now call these plugins. An app lets someone use your product inside ChatGPT once they’ve chosen it; it doesn’t make ChatGPT recommend you in an answer. Build one if your product is genuinely useful in a conversation (search over data, quotes, scheduling), and treat it as a product, not as AI visibility. The other routes are in how to get your business on ChatGPT.
FAQ
How does ChatGPT decide which software to recommend?+
When it searches, it retrieves pages that answer the buyer’s question (comparison articles, “best X software” lists, vendor pages, forum threads) and recommends what they name, with reasons drawn from them. Without search, it relies on what it learned in training. Being described consistently and specifically across those sources is what gets you named.
Do G2 and Capterra reviews help with ChatGPT?+
Less than many teams assume, on 2026 data. In GetIntel’s index of 126 categories, G2, Capterra and TrustRadius together were about 5% of ChatGPT and Gemini citations. Reviews still matter to buyers and to list authors who rank tools, so keep them current, but don’t treat them as the main AI lever.
Should we publish comparison and alternatives pages?+
Yes, if they’re honest. Buyers ask assistants “X vs Y” and “alternatives to X”, and the pages that answer get read. Ahrefs found a brand’s own list appeared in 34% of ChatGPT software answers where that brand was recommended. Compare fairly, say who each tool suits, and keep prices current.
Does publishing an app in ChatGPT get our software recommended?+
Not by itself. An app (OpenAI now calls them plugins) lets people use your product inside ChatGPT once they choose it; recommendations come from the web sources ChatGPT reads. Build one if users would genuinely use your product in a conversation.
Sources
- ChatGPT and Gemini disagree on the best software in a third of categories, and the review sites have been skipped (GetIntel AI Software Index) — The Next Web (2026-08-12)
- Do self-promotional “best” lists boost ChatGPT visibility? (26,283 source URLs) — vendor study — Ahrefs (2025-12-04)
- Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts) — vendor study — Ahrefs (2026-04-15)
- Brand mentions vs AI visibility (75,000 brands, correlational) — vendor study — Ahrefs (2025-12-12)
- Why calling yourself “the best” could cost you AI recommendations (Ahrefs Brand Radar data) — Lily Ray (2026-06-17)
- The rise of the AI crawler (JavaScript rendering) — Vercel (2024-12-17)
- Developers can now submit apps to ChatGPT — OpenAI (2025-12-17)
Run it on your domain
See this lens as a finding, not a lecture.
Free technical GEO audit. Evidence-traced scores and an LLM-ready export. See your score with no login.
Audit your website
