Each AI system finds sources through its own index. Get the baseline framework right first, make sure you are in the index behind the platform your audience uses, then tune from your own monitoring data.
By Christopher Foster-McBride, Founder, AI Knowledge Signal · Reviewed 2026-11-03
Why it matters
Different AI systems find sources in different places. Google AI Overviews and AI Mode draw on the Google Search index; Copilot draws on Bing; ChatGPT search uses OpenAI’s OAI-SearchBot index plus third-party search providers; Perplexity runs its own crawler; Claude’s web search uses Anthropic’s search crawler and a search provider. If you are missing from the index behind a platform, no amount of content tuning will get you cited there.
The discipline is sequenced: a page built to the baseline framework is eligible everywhere; then confirm you are indexed by each platform that matters; and only then tune, using what your own citation log (Phase 6.3) shows. Published claims about what each engine “prefers” change quickly and are rarely replicated — your own data beats a generic table.
Bing now reports your grounding footprint by topic [Beta]
The AI Performance report in Bing Webmaster Tools has a Topics Portfolio tab that clusters the grounding queries where your site appears into themes, with demand volume and competitive intensity per topic, personalised to your site. That makes it a gap diagnostic, not just a presence count: it shows themes with demand where your coverage is thin. It covers Bing AI grounding only, so treat a topic gap as a hypothesis to test on your other monitored platforms. Export it quarterly and record the export date in your citation log (Phase 6.3).
Baseline first, then deliberate tuning to the platform that matters most:
The page meets the baseline framework (Phases 1–5).
You are in the index behind each priority platform (table below) — crawler allowed, pages indexed.
One priority platform is chosen — the one your audience actually uses.
Citation presence is monitored on each platform separately (Phase 6.3).
Tuning decisions come from your own monitoring data, not a generic table.
Step by step
Get the baseline framework right first
Before any platform tuning, make sure the page satisfies Phases 1–5. A baseline-compliant page performs adequately on all platforms; tuning is an add-on, not a substitute.
Identify the platform your audience uses
Pick the single platform where your audience actually asks questions — don't target all simultaneously from day one. Prioritisation beats thin coverage everywhere.
Get into that platform's index
Use the table below: allow the platform’s crawler (Phase 1.1), and use its webmaster tools — Google Search Console for Google’s AI features; Bing Webmaster Tools and IndexNow for Copilot. Then confirm with a few test prompts that your pages can be retrieved at all.
Monitor each platform separately
Track citation presence per platform — patterns differ significantly, so a single blended number hides what's working. This is the per-system citation log from Phase 6.3.
Tune from your own data, then expand
Look at which of your pages and which third-party sources each engine cites for your prompts, and close the gaps you see (e.g. an engine that keeps citing a Wikipedia article or a YouTube explainer you don’t have). Then extend to the next platform.
Worked examples
1 Where each platform finds sources
Documented retrieval sources and the controls you have. Re-check each vendor’s documentation at every quarterly review.
Platform
Retrieves from
What you can do
Google AI Overviews, AI Mode, Gemini
Google Search index; several related “fan-out” queries per question
Normal SEO; Search Console; cover sub-questions on one strong page. Blocking Google-Extended removes Gemini-app grounding
Microsoft Copilot
Bing index
Bing Webmaster Tools (AI Performance, including the Topics Portfolio beta); IndexNow
ChatGPT search (OpenAI)
OAI-SearchBot index + search partners; ChatGPT-User
Allow both agents
Perplexity
PerplexityBot index + Perplexity-User fetches
Allow both agents
Claude (Anthropic)
Claude-SearchBot + a search provider; Claude-User
Allow both agents
2 Before / after: scattershot vs. prioritised
Before
Trying to optimise for every AI system at once from day one, with no per-platform monitoring — effort spread thin and no clear signal of what works.
After
Baseline framework applied everywhere; Perplexity chosen as the priority (audience uses it most); PerplexityBot confirmed allowed; citation log shows it cites a competitor’s comparison table, so a comparison page is added; presence monitored per platform.
3 B2B: LinkedIn as a source in its own right
For professional queries, AI answers often cite LinkedIn pages directly rather than your site. LinkedIn’s October 2026 guide reports it is the most-cited domain for those queries (Profound) and that about 75% of its citations come from individual member profiles, not Company Pages (Meltwater). These are vendor-reported figures, and Meltwater’s wider data ranks LinkedIn second to YouTube across all queries. Check your own log before you treat LinkedIn as the priority.
Format
Role in AI answers (LinkedIn’s data)
What to do
Articles (long-form)
~60% of LinkedIn content citations; more likely to be the primary source
Make it the canonical piece on a topic. The title becomes the URL, so write it as the question your audience asks
Posts (short-form)
~40% of citations; often a route to a longer asset
One idea per post, pointing back to the article. The first line becomes the URL slug and can’t be changed later, so lead with the key terms and skip emojis and hashtags
Member profiles
Most citations come from individuals, not Company Pages
Publish under named experts; link profile ↔ site (Phase 4.1)
Visibility settings
A profile or post AI can’t reach can’t be cited
Profile public, posts visible, comments on. Ask an assistant to open the URL to confirm it can
Industry studies (for example Goodie’s AEO Periodic Table) report that engines differ in which third-party sources they cite — Wikipedia, YouTube, Reddit and review sites in different proportions. Use that as a prompt to check your own log, not as a rule.
Common mistakes
Mistake
Fix
Platform tuning before the baseline
Get Phases 1–5 right first; tuning is second-order.
Tuning content for a platform whose index you’re not in
Check crawler access and indexing first (Phase 1.1, Search Console, Bing Webmaster Tools).
Targeting all platforms at once
Prioritise the one your audience uses most.
One blended citation metric
Monitor each platform separately — patterns differ.
Treating a published "what each engine prefers" table as fact
Use it as a hypothesis and test it against your own citation log.
Limits of this advice
Vendors change retrieval partners and crawler names; re-check the table quarterly.
Published per-engine “preference” studies are directional and rarely replicated; your own monitoring is the evidence that counts.
Platform-published guidance (such as LinkedIn’s guide on LinkedIn) is written by the platform about itself. Use its format advice, but confirm the citation claims against your own log.
Checklist
Page meets the baseline framework (Phases 1–5)
Crawler access and indexing confirmed for each priority platform
One priority platform chosen (where the audience is)
Citation presence monitored on each platform separately
