How to get your SaaS recommended by ChatGPT, Claude and Perplexity

How assistants choose which software to name

Two paths: model memory and live search

An assistant can answer from what the model learned in training, or it can search the web first and build the answer from what it retrieves. Which path it takes varies by assistant, by question and even by run: one configuration reviewed in a July 2026 survey of 45 studies didn’t trigger a web search in 57.8% of ChatGPT repetitions.^[1] Live search is the path you can influence quickly. Model memory changes only when the model is updated.

Each assistant reads a different index

So ranking well on Google helps with Google’s AI features but tells you little about ChatGPT, Claude or Perplexity.

Being cited is not the same as being recommended

An assistant can link to your page as a source and still recommend a competitor, or name you without linking anywhere. Peec found that ChatGPT’s Deep Research agent reads far more pages than it cites, so pages can shape an answer without ever appearing as a citation.^[6] Track both, separately.

Answers change from run to run

In SparkToro and Gumshoe’s study of 2,961 runs, ChatGPT and Google’s AI search returned the same list of brands less than 1% of the time, and the same list in the same order less than 0.1% of the time.^[7] Presence was steadier than position: in one category, the leading brands appeared in 55–77% of responses. Getting recommended therefore means raising how often you’re named, not winning a slot.

Contrary evidence: the same study found narrow categories were more consistent than broad ones, so a dominant leader in a tight B2B niche may be named almost every time. “Everything is random” overstates it.

Step 1: Make sure assistants can read your site

Separate search bots from training bots in robots.txt

OpenAI and Anthropic each run separate crawlers for search, for training and for fetches a user asks for, and each can be allowed or blocked on its own.^[2][4] Many SaaS sites block “AI bots” as one group and disappear from AI search as a side effect. If you don’t want your content used for training, block GPTBot and ClaudeBot, and still allow OAI-SearchBot, Claude-SearchBot and PerplexityBot.

Check your CDN and firewall, not just robots.txt

Since 1 July 2025, Cloudflare blocks AI crawlers by default on new domains.^[8] Firewall rules and bot protection can block a crawler that robots.txt allows, and nothing in your analytics will tell you. The reverse is also true: robots.txt is a request, and Cloudflare has reported a crawler ignoring it.^[9]

Put pricing, features and docs in the HTML

Vercel and MERJ found that the crawlers from OpenAI, Anthropic, Meta, ByteDance and Perplexity didn’t execute JavaScript: ChatGPT’s and Claude’s crawlers downloaded JavaScript files but never ran them.^[10] If your pricing table or feature list only appears after a React app loads, those crawlers see an empty page.

Contrary evidence: Googlebot does render JavaScript, and some assistants retrieve through search indexes whose crawlers render too, so client-side rendering hurts some paths, not all. The Vercel data is from December 2024; check your own server logs.

Check it yourself

Our free AI crawler checker tests your robots.txt rules for each AI crawler, fetches the page with their user agents, and shows what’s in the HTML before JavaScript runs.

Step 2: Publish pages that answer buyer questions in the first lines

Comparison, alternatives and use-case pages

B2B buyers ask assistants for shortlists (“best SOC 2 automation tool for a 30-person startup”), for alternatives (“Vanta is too expensive, what else is there?”) and head-to-heads (“Vanta vs Drata vs Secureframe”). If no page anywhere states plainly how you compare, an assistant has nothing to retrieve. Publish honest comparison and “alternatives to [leader]” pages, and a page per main use case.

Answer first, then the detail

Put the direct answer in the first paragraph, as text, then support it. You don’t need to cut pages into tiny chunks: Google’s guide says there’s “no requirement to break your content into tiny pieces”.^[5] Be careful with wholesale rewrites aimed at AI: in one benchmark reviewed in the GEO survey, body-only rewrites raised citation once a page was selected but cut its top-10 presence after reranking by 16%.^[1]

Specific, dated, checkable facts

A line like “Type II report in N weeks, from $X a year, as of October 2026” gives an assistant something to quote. “Lightning-fast compliance” doesn’t. Keep facts consistent across your site, docs, review profiles and directory listings.

Keep key pages fresh

In SE Ranking’s study, pages updated within the last 3 months averaged 6.0 ChatGPT citations against 3.6 for older pages.^[11] That’s a correlation: well-maintained sites tend to be stronger in other ways too. Update pages when facts change, not to game a date.

Step 3: Be present where assistants look for corroboration

Review sites

G2 is among the ten most-cited domains across AI engines in Peec’s dataset of about 30 million cited sources.^[12] In G2’s own survey of 1,076 B2B software buyers, review-site citations were the top trust signal in AI answers.^[13] Both sources have a commercial interest in this conclusion; it’s still the most direct evidence available.

Reddit, YouTube and LinkedIn: honest participation only

Reddit, YouTube and LinkedIn top Peec’s cross-industry list,^[12] but industries have their own dominant sources, so map yours before investing. Answer real questions as yourself. Google’s guide says seeking inauthentic mentions is unlikely to help,^[5] and planted posts are a reputational risk when they’re found.

Brand mentions correlate more than backlinks

In Ahrefs’ study of 75,000 brands, branded web mentions correlated with AI Overview visibility at 0.664, against 0.218 for backlinks.^[14] It’s a correlation, and it excluded brands with no mentions at all, but it points the same way as everything above: be talked about accurately in the places buyers and assistants read.

Step 4: Structured data and llms.txt: what the evidence says

Schema markup

Ahrefs compared 1,885 pages that added JSON-LD with 4,000 matched pages that didn’t. Over 30 days, AI Mode and ChatGPT citations moved by +2.4% and +2.2%, which it couldn’t distinguish from zero, and AI Overview citations fell 4.6%. Its conclusion: it “can’t tell whether the schema did a tiny bit of good or nothing at all”.^[15] In SE Ranking’s data, pages with FAQ schema averaged slightly fewer ChatGPT citations (3.6) than pages without (4.2).^[11] Google says structured data isn’t required for its generative AI features.^[5]

Contrary evidence: cited pages were about three times more likely to carry JSON-LD in the same Ahrefs data (a correlation), and critics argue schema’s job is entity clarity, not citation lift.^[16] We keep accurate Organization, Product and FAQ markup that matches the visible page, for clarity and rich results, not as an AI lever.

llms.txt

Google says its Search doesn’t use llms.txt,^[5] and SE Ranking found its effect on ChatGPT citations negligible.^[11] It takes minutes and does no harm, so add one if you like, but fix crawler access and server rendering first.

Step 5: Measure before and after

Write down 15–25 questions buyers actually ask, freeze the list, and run each one several times per assistant in fresh sessions. Report “named in X of N” with the sample size, plus who was named instead and which sources were cited. Then make changes and re-measure the identical set. Our methodology sets out the full protocol, and Measuring AI visibility explains the statistics.

One screenshot proves nothing: given the run-to-run variation above, a single answer can show you in or out by chance.

What doesn’t work, or isn’t proven

• “GEO boosts visibility by 40%.” The figure comes from one metric in a lab setup where documents were already in the model’s context. The GEO survey rates it lowest confidence and found no technique with a stable, cross-platform causal effect.^[1]
• Keyword stuffing and AI-only rewrites. See the rewrite result in Step 2.
• Hidden text or separate pages served only to AI crawlers. It’s cloaking, and it risks your search rankings.
• Fake reviews and planted forum posts. Against platform rules, and unlikely to help.^[5]
• Paid “guaranteed placement” in AI answers. No AI company sells placement inside its organic answers.

Contrary evidence: the same survey includes one log-based study that found a measurable ChatGPT referral effect (multiplier 1.82, 95% interval 1.31–2.54), so “nothing works” is too strong as well.^[1]

How long it takes

This is reasoning, not a measured result. Answers built from live search can change once your pages are recrawled and re-ranked, often within weeks. Answers from model memory change only when the model is retrained or updated, which can take months. Re-measure on the same prompt set to see which you’re dealing with.

What we’d do first for a seed to Series B SaaS

1. Check robots.txt and CDN rules for OAI-SearchBot, Claude-SearchBot and PerplexityBot (free checker).
2. Make sure the home, pricing, features and comparison pages show their key facts in the HTML without JavaScript.
3. Publish one honest “alternatives to [category leader]” page and one comparison page, each answering in the first paragraph.
4. Bring your G2 profile and the two or three third-party sources assistants cite in your category up to date and accurate.
5. Measure 15–25 buyer prompts with repeated runs before changing anything, then re-measure.

The free snapshot does a small version of step 5 for you: 3 buyer questions, 3 assistants, 3 runs each. The audit covers all five.

Questions

Related

• AEO vs SEO for B2B SaaS: what differs and what doesn’t
• Measuring AI visibility: why “rank in ChatGPT” misleads
• How AEOLedger measures AI visibility (methodology)

Sources

All links checked on 17–18 October 2026. Vendor interests are noted for each source.

1. Martinez, critical survey of 45 generative engine optimization studies (arXiv 2607.14035), submitted 15 July 2026. Single-author preprint reviewing 45 studies; not peer reviewed. Some figures quoted here come from PPC Land’s summary (ppc.land), not the paper itself.
2. OpenAI documentation, OpenAI crawlers and user agents, checked 17 October 2026. Primary documentation.
3. Xponent21, analysis of Claude web search providers (Brave, turbopuffer), 2 July 2026. Agency analysis of Anthropic’s published subprocessor list; overlap figures are vendor-reported and directional.
4. Anthropic help centre, Anthropic’s crawlers and how to control them, checked 17 October 2026. Primary documentation.
5. Google Search Central, “Optimizing your website for generative AI features on Google Search”, published 15 May 2026, last updated 10 July 2026. Covers Google AI Overviews and AI Mode only.
6. Peec AI research, pages read vs cited in ChatGPT Deep Research, June 2026. Vendor. No sample size published.
7. Search Engine Journal, coverage of SparkToro and Gumshoe’s AI recommendation consistency study, 30 January 2026. Coverage of SparkToro and Gumshoe’s study. Gumshoe sells AI visibility tracking. Not peer reviewed; consumer categories, not B2B SaaS. Original: sparktoro.com.
8. MIT Technology Review, report on Cloudflare blocking AI crawlers by default, 1 July 2025. News report.
9. Cloudflare blog, report on undeclared Perplexity crawlers ignoring no-crawl directives, August 2025. Cloudflare sells bot management.
10. Vercel and MERJ, “The rise of the AI crawler”, 17 December 2024. Vercel sells hosting. One network’s traffic; nearly two years old, so verify on your own logs.
11. Search Engine Journal, coverage of SE Ranking’s study of ChatGPT citation factors, 26 November 2025. Vendor (SE Ranking sells SEO tools). 216,524 pages across 129,000 domains; correlational; ChatGPT version not stated.
12. Peec AI research, most-cited domains in AI search, March 2026 data. Vendor (Peec sells AI visibility tracking). About 30 million cited sources, US only; Claude not covered.
13. G2, “The Answer Economy”, 15 April 2026. G2 sells review-site presence. Survey of 1,076 B2B software buyers, March 2026; self-reported behaviour.
14. Ahrefs, brand visibility factors in AI Overviews, May 2025. Vendor. 75,000 brands; Spearman correlations only; excluded brands with zero mentions and DR below 40.
15. Ahrefs, difference-in-differences study of schema markup and AI citations, 11 May 2026. Vendor. Difference-in-differences: 1,885 pages that added JSON-LD vs 4,000 matched controls, 30 days; already heavily cited pages only.
16. iloveseo.net, critique of the Ahrefs schema study, 2026, checked 18 October 2026. Commentary, not new data.

How to cite and changes

AEOLedger. “How to get your SaaS recommended by ChatGPT, Claude and Perplexity.” Guide v1.0, 18 October 2026. https://aeoledger.com/guides/get-recommended-by-chatgpt-claude-perplexity/

• v1.0, 18 October 2026: first published.

Spotted an error or a better source? Write to hello@aeoledger.com. Corrections are dated here.