AI Search vs SEO: Competing Channels or One System?

Framing "AI search vs SEO" as a choice gets the mechanics wrong. An AI search engine usually runs a search first, then writes an answer from the pages it found. Those pages come from an index that a crawler built. That is the same machinery SEO has always worked on.

So think of AI answers as a pipeline with gates:

• Access: can the engine's crawler fetch and read your page?
• Index: did that engine's index keep it, and is it allowed to show a snippet?
• Retrieval: did the page come back for the queries the engine ran?
• Generation: did the model use and cite it in the answer?

SEO work lives mostly in the first two gates and partly in the third. The fourth belongs to the model. That is why SEO health feeds AI answers but does not determine them. If you want the wider comparison of the two kinds of engine, see answer engines vs search engines.

Google is explicit about its own surfaces. Its AI features documentation says a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode. It adds that there are "no additional technical requirements" and that indexing and serving "isn't guaranteed."

Other engines use other indexes. As of October 2026, the documented picture looks like this:

Two details change how you read this table.

First, engines don't search for the prompt the user typed. OpenAI says ChatGPT search rewrites the query into one or more targeted queries. Google may fan a question out into several related searches. You never see those queries, so ranking for one keyword is a weak proxy for being retrieved.

Second, retrieval is optional. ChatGPT decides per question whether to search at all. When it doesn't, the answer comes from what the model learned before its knowledge cutoff. Nothing you publish this quarter reaches that answer.

Ahrefs found that 37.9% of URLs cited in AI Overviews also appeared in the first 10 results for the query, in a study published in March 2026. The organic-only figure was 37.1%. Roughly 31% of cited URLs ranked in positions 11 to 100, and roughly 31% ranked beyond 100.

What it means for you:

• Ranking is associated with citation, not required for it. Most cited URLs were not in the top 10.
• Ranking is not sufficient either. Plenty of top-10 pages are not cited.
• It is one engine. The figure covers Google AI Overviews, not ChatGPT, Perplexity or Copilot.

Read it as a snapshot. Ahrefs changed its parsing method between its studies, so earlier figures are not comparable and do not form a trend. Ahrefs also matched at the URL level, which can undercount large sites cited through a page that doesn't rank. Other vendors report different numbers with different definitions, so treat any single overlap figure with care.

The practical reading: classic SEO keeps you in the candidate pool on Google. It doesn't decide who gets quoted. For what moves the rest, see SEO for AI search.

OtterlyAI ran a public test in 2026: two new domains, each with roughly 1,000 AI-generated posts. Google deindexed both, one after about 14 days and the other after about 90, with no manual action reported. Bing kept both sites indexed.

The AI citations followed the indexes. Otterly tracked 3,934 citations for one site, and 2,718 of them came from Microsoft Copilot, which grounds on Bing. Google's AI surfaces almost never cited the sites once Google had dropped them. ChatGPT cited them effectively zero times, even while they ranked in Bing and ChatGPT's crawlers visited.

Three lessons, with the caveat that this is one pair of sites run by a vendor:

• An engine can't cite what its index no longer holds.
• Being indexed is necessary but not enough. "Rank in Bing to appear in ChatGPT" did not hold here.
• Engines diverge sharply, so a single blended number hides what is happening. Our comparison of one score vs per-platform scores covers why.

In your control

• Rendering. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. Content that only appears after JavaScript runs may be invisible to them. Googlebot does render, so this gap often shows up outside Google.
• Per-bot robots.txt rules. OpenAI states that its crawler settings are independent. GPTBot covers training. OAI-SearchBot covers ChatGPT search.
• Network reachability. A firewall or CDN can block a crawler that robots.txt allows. OpenAI's help centre says eligibility needs both robots.txt permission and a host that allows its published search crawler IPs.
• Index and snippet eligibility. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers on Google and Bing.
• Clear facts on fetchable pages. Stating a price, a number or a one-line definition in plain HTML gives an engine a sentence it can quote.

Not in your control

• Which queries the engine generates from a user's prompt.
• Where your page lands in the model's list of retrieved sources.
• Whether the model searches at all, or answers from memory.
• Which retrieved page it cites.

The honest summary: you can remove blocks. You cannot force a citation.

Work from the gates you control toward the ones you don't. A small team can do the first pass in an afternoon, and our one-afternoon AI visibility check walks through it.

1. Check crawler access per bot. Read your robots.txt for Googlebot, bingbot, OAI-SearchBot, PerplexityBot and Claude-SearchBot. A blanket AI block often removes ChatGPT search eligibility by accident. If you want training opt-out without losing search, the rules look like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

2. Check the firewall, not just the file. Request a page as each crawler and look for 403 responses or challenge pages. A crawl test such as 99Visibility's AI Crawler Checker shows which crawlers your firewall turns away.
3. Diff raw HTML against the rendered page. Make sure prices, product names and key claims exist before JavaScript runs.
4. Audit suppression tags. Look for noindex, nosnippet and data-nosnippet inherited from templates.
5. Check each index separately. Use Google Search Console and Bing Webmaster Tools. Health in one says nothing about the other.
6. Put the facts that matter in plain sentences. Pricing, what you do, who you serve, where you operate.
7. See what the engines actually say. The free checker on this site shows whether AI engines name you for three buyer questions. Treat it as a snapshot, then fix wrong facts at the source.

To test a single page's SEO basics, the free AI SEO checker is a quick start.

• Treating AI search as a separate channel that needs new tricks. Google's AI optimization guide says you don't need special markup, AI text files or a special writing style to appear in its generative features.
• Assuming top-10 rankings guarantee citations. The Ahrefs figure shows most cited URLs sat outside the top 10.
• Allowing GPTBot to "get found by ChatGPT." GPTBot governs training. ChatGPT search uses OAI-SearchBot.
• Buying markup as an AI lever. Google says structured data isn't required for generative AI search, and llms.txt is ignored by Google Search. Existing markup does no harm, but it is not a citation lever.
• Reading one blended score. Engines draw on different indexes and can behave in opposite ways for the same site.
• Mass-publishing thin pages. Google's spam policies treat scaled content without user value as abuse, and the removal can arrive silently.
• Declaring SEO dead. AI answers still rest on indexes. For the longer argument, see whether AI will replace SEO.