Can AI Do SEO? What It Can and Can't Handle

AI is good at SEO work that is repetitive, pattern-based and easy to check. That covers more of the job than most people expect.

• Research grunt work. Grouping hundreds of keywords by intent, summarising competitor page structures, drafting topic maps.
• Content production. Outlines, briefs, first drafts, title and meta description variants, alt text.
• Technical support. Writing regex for log files, drafting redirect maps, generating JSON-LD, explaining a crawl report.
• Analysis. Spotting patterns in Search Console exports, flagging pages with thin or duplicated copy.

The common thread: AI produces a draft that a person can verify quickly. When the output is cheap to check, AI saves real time. When the output is expensive to check (a claim about your pricing, a statement about a competitor), AI shifts the work rather than removing it.

If you want the mechanics of wiring these tasks together, see how to automate SEO tasks with AI and the practical notes on using ChatGPT for SEO work.

Use this as a split of labour, not a ranking of importance.

Two rows deserve a caution. AI can write schema markup correctly, but Google says structured data isn't required for generative AI search, and Ahrefs' matched-control study found AI Overview citations fell 4.6% after pages added schema. Treat schema as ordinary hygiene, not a route into AI answers. And AI agents that act on your site carry their own risks, covered in what AI SEO agents do and where they fail.

Google's position is plain. Its guidance on AI content says: "Appropriate use of AI or automation is not against our guidelines." Its spam policies define scaled content abuse as "when many pages are generated for the primary purpose of manipulating search rankings and not helping users," and the first example is using generative AI to generate many pages without adding value.

Otterly tested the volume shortcut in 2026. It launched two new domains with roughly 1,000 AI-generated posts each. Otterly reports the site with a fictional expert author box lasted about 14 days in Google's index, and the site with no author lasted about 90. Neither received a manual action in Search Console. The removal was silent.

The AI side was different. Otterly tracked 3,934 AI citations for one site, and 2,718 of them came from Microsoft Copilot, which grounds on Bing. Bing never dropped either site. Google's AI surfaces almost never cited them after Google removed the pages. The Bing and Google index split explains why one engine can keep citing what another has dropped.

The test has limits: one site per arm, no human-written control, a low-stakes topic, and a vendor ran it. It still points one way. Using AI to write is fine. Using AI to publish at volume without adding value puts your Google presence at risk, and nothing warns you first.

Verify facts

A language model writes what is plausible, not what is checked. It will state a price, a founding year or a feature list with the same confidence whether it is right or wrong. AI platforms also get specific facts wrong about specific brands in their answers. Your defence starts on your own site: state each key fact once, consistently, in plain HTML, and have a person check every AI draft against your records.

Earn mentions

When engines cite other sites instead of yours, being named on those sites is the route into the answer. The evidence shows a strong association here, not a guaranteed effect. AI can draft an outreach email. It can't make you worth mentioning, and it can't build the relationship. Some reference sites have strict rules too; read the Wikipedia editing rules for brands before anyone touches an article about you.

Control retrieval

AI answers pass through four stages. The model's built-in knowledge (parametric) is frozen at a training cutoff. Access decides whether a crawler can fetch your page. Retrieval decides whether your page enters the model's context for a given query. Generation decides what gets used and cited. You control access. You partly influence retrieval through ordinary search visibility. You control nothing else. As of October 2026, ChatGPT decides per query whether to search at all, and OpenAI says it rewrites a user's query into one or more targeted queries before searching.

A popular pitch says: let AI rewrite your pages in an "AI-friendly" style and engines will cite you more. The best evidence says otherwise.

• C-SEO Bench (NeurIPS 2025) found that "most current C-SEO methods are not only largely ineffective but also frequently have a negative impact on document ranking."
• A July 2026 critical survey of 45 GEO studies concluded that no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability.
• Google's own optimization guide says you don't need to write in a specific way for generative AI search, and rejects chunking content for LLMs as a requirement.

There is one positive signal for on-page content. Stating a specific fact (a price, a number, a one-line definition) on a page that can be fetched gives an engine a sentence it can quote. That is a reason to write clearly about what you sell. It is not a reason to run every page through a rewriting tool.

The levers that hold up are access levers. They work by removing blocks, and you can check each one.

In your control:

• Rendering. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. If your pricing or product copy only appears after JavaScript runs, those crawlers may not see it.
• Per-bot robots.txt rules. OpenAI's crawlers are independent: GPTBot covers training, OAI-SearchBot covers ChatGPT search.
• Network reachability. A firewall or CDN can block a crawler that robots.txt allows.
• Index and snippet eligibility. Google says a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out.

A robots.txt that keeps ChatGPT search access while opting out of training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Not in your control: whether a query triggers a search, which pages an engine retrieves, where your page lands in the model's context, and what the model finally says. No AI tool or agency controls these either. Anyone promising a citation or a score is promising something they can't deliver.

1. Check access first. Confirm your robots.txt lets the search crawlers in (OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot). Then confirm your firewall doesn't turn them away. 99Visibility's AI Crawler Checker requests a page as each AI crawler and shows which ones get blocked.
2. Check rendering. Load key pages with JavaScript off. If prices, product names or core claims vanish, fix that with server-side rendering.
3. Check indexing per engine. Look at Google Search Console and Bing Webmaster Tools separately. A healthy Google index tells you nothing about Bing, and the reverse is true. See whether you need a Google ranking to show up in ChatGPT.
4. Fix your facts. Put pricing, plans and key product facts in one consistent, fetchable place. Have a person verify them.
5. Use AI for production. Briefs, drafts, meta tags, clustering. Review every factual claim before publishing.
6. Work on mentions. Find the sites engines cite for your buyer questions and earn a place on them honestly.
7. Measure each engine separately. The free checker on this site shows whether AI engines name you for up to three buyer questions, as a snapshot.

If the access and rendering steps feel out of reach, decide whether to bring in expert help or keep it in-house.

• Mass-publishing AI pages. Otterly's test shows Google can remove a site with no warning in Search Console.
• Publishing unchecked facts. A wrong price on your own site gives engines a wrong sentence to quote.
• Allowing GPTBot to get into ChatGPT search. GPTBot governs training. OAI-SearchBot governs search.
• Blanket-blocking all AI bots. This can silently remove ChatGPT search eligibility along with training use.
• Buying schema, llms.txt or E-E-A-T fixes as AI levers. Google says structured data isn't required for generative AI search, Google Search ignores llms.txt, and E-E-A-T isn't a specific ranking factor or a score.
• Trusting one blended score. In Otterly's test one engine cited a site thousands of times while four barely did. An average would have hidden that.
• Injecting hidden instructions for AI crawlers. Hidden text and prompt injection exist and research shows they can sway some systems. They are manipulation, they put your reputation and index status at risk, and they have no place in a legitimate plan.

Will Google penalize content written with AI? Not for being AI-written. Google's guidance says appropriate use of AI or automation is not against its guidelines. Google acts against scaled content abuse, meaning many pages produced to manipulate rankings without helping users. The risk comes from volume without value. In Otterly's experiment, two sites with roughly 1,000 AI posts each were removed from Google's index with no manual action reported.

Can AI tools tell me my rank inside AI answers? No tool can verify one. As of October 2026, Google's Search Console report for AI Overviews and AI Mode shows impressions only, with no position data. OpenAI says only that ChatGPT ranks results using multiple factors. What a tool can honestly report is whether an engine named or cited you for a specific question, engine by engine, with the date it checked.

Should I use AI to rewrite my pages so ChatGPT cites them? The evidence says it is a poor use of time. C-SEO Bench found most conversational SEO rewriting methods largely ineffective and often harmful. Google says you don't need to write in a special way for generative AI search. A better use of AI is drafting clear pages that state your key facts plainly, then having a person check them.

Can an AI SEO agent run my SEO without supervision? It can run repetitive tasks, but unsupervised agents can publish wrong facts, change robots.txt or meta robots tags, or produce pages at a volume Google treats as scaled content abuse. Keep a person approving anything that goes live, anything that changes crawler access, and every claim about your prices or products.