AI-powered SEO in 2026: rank on Google, get cited by ChatGPT

Two audiences, not one#

Five years ago, "doing SEO" meant optimising for Google. In 2026 it means optimising for two different retrieval systems with two different incentives: the Google search index and the new answer engines — ChatGPT search, Perplexity, Claude with web access, Gemini, Copilot, and whatever ships next quarter.

Google still drives the majority of clicks for most sites. But an increasing share of discovery — especially for B2B and technical products — now happens inside a chat window where the model picks which sources to cite. If your page is not in that citation set, you are invisible to a growing slice of your market.

The good news: the two systems agree more than they disagree. The playbook that gets you cited by an answer engine is also the playbook that ranks well on Google. The difference is mostly in emphasis.

What changed in the last two years#

Three things have shifted, and each one changes the writing.

AI Overviews pushed answer-first content above the fold. Google's AI Overviews now sit above the organic blue links for a large share of informational queries. The content that gets pulled into the Overview is content that already answered the question cleanly, with structured data and clear factual statements the model could lift.

Answer engines cite aggressively when given clean signals. ChatGPT search, Perplexity, and Claude with web browsing all share a habit: they prefer sources that make claims explicitly, with author, date, and source URL visible in the markup. A page that buries the answer in narrative is harder to cite than a page that opens with the answer and then justifies it.

Entity consistency across the web matters more than ever. If your company name, address, founders, and product descriptions say different things on your site, your LinkedIn, your Crunchbase entry, and your Wikipedia stub, answer engines hesitate to cite you. Models are suspicious of inconsistent entities.

The overlap: what works for both#

Most of the fundamentals are unchanged. Both Google and the answer engines reward:

• Structured data. Schema.org markup for your organisation, your articles, your products, your FAQ. Not because Google needs it to rank — it does not — but because it removes ambiguity the model has to resolve by guessing.
• FAQ sections on commercial and informational pages. Both systems love Q&A markup. A clean FAQ answers the exact phrasings real users type, and both systems will lift it.
• Author and date metadata visible on the page. Both systems discount undated content when they have a choice.
• Clear, factual claims in the first two sentences. Not hook copy, not vibes — the answer.
• Original data, primary sources, and named experts. Generic content has been discounted for years; the discount is now harsher.

What is different: optimising for citations#

A few practices have become more important now that answer engines are a meaningful share of traffic.

Treat citation-worthy statements as a first-class output. A claim like "the average response time for our support team is under two hours, based on the last 90 days of ticket data" is more citable than "we pride ourselves on fast support." The first is a sentence a model can quote. The second is sentiment.

Maintain a clean llms.txt. The convention is simple: a plain-text file at the root of your site that lists what your company is, what you do, and which pages are authoritative on which topics. Answer engines read it. Keep it short and factual.

Keep entity info consistent. Same company name, same address, same founders, same product descriptions everywhere they appear — your site, your profiles, your press, your schema. Inconsistency is the single most common reason answer engines hesitate to cite a business.

Earn citations in places models trust. Reviews on G2, Capterra, and Trustpilot. Coverage in trade publications. Listings in authoritative directories. Answer engines lean on the same authority signals Google does, just with different weights.

AI-assisted content pipelines done right#

I do not pretend AI-assisted content is a bad idea. It is the only way most small teams can publish consistently. The trick is the workflow.

A pipeline I am comfortable with: an analyst defines the topic, the search intent, and the ten questions the page needs to answer. A model produces a first draft that hits all ten. A human editor verifies every factual claim, adds the original data and the named experience, and rewrites the opening so it is not generically written. Then it ships on a schedule — weekly or biweekly, not in bursts.

What I will not do, and what I tell clients not to do: publish the model's first draft untouched, stuff keywords into AI-generated text, or generate hundreds of pages of thin content and hope one of them ranks. All three still work less than the marketing says, and the reputational cost when you get caught is real.

What to measure in 2026#

Rank tracking alone is no longer enough. A useful dashboard in 2026 has at least four numbers:

• Search Console impressions and clicks. Still the source of truth for Google traffic. Watch the impression curve, not just the click curve — AI Overviews steal clicks but not impressions.
• Citation rate in answer engines. Sample the fifty questions your customers actually ask. Ask ChatGPT search, Perplexity, and Claude. Count how often you are cited, and against which competitors.
• Indexed pages and crawl health. Both systems still need to find and parse your content. A page that is not in the index is not cited.
• Conversion to qualified next step. None of this matters if the traffic does not become pipeline.

If you want to set this up properly for your site, the AI-powered SEO & content service walks through the engagement. If you want to talk through your specific situation first, send me a short note and I will tell you whether you are looking at a one-month tune-up or a six-month program.

---

Frequently asked questions#

Is traditional SEO dead because of AI Overviews?#

No, but it has changed. AI Overviews reduce clicks on informational queries — the "what is" and "how to" searches — much more than on commercial and transactional queries. If your business depends on informational traffic, the strategy has to shift toward citations and brand authority. If your business depends on commercial traffic, the work is closer to classic SEO plus entity hygiene.

How long does it take to get cited by ChatGPT or Perplexity?#

For a brand that already has consistent entity data, clean schema, and a few authoritative mentions on the open web, first citations tend to appear within a couple of months of focused work. For a brand new site with no existing footprint, plan on three to six months of patient link-earning and content publishing. There is no shortcut — the answer engines reward the same authority signals Google does, just with slightly different weights.

Should I be using AI to write all my content?#

Use AI to draft faster, not to publish faster. The pages that rank and get cited are the ones a human verified and improved. The pages that get penalised or ignored are the ones where the model's output shipped untouched. The win is in the human review step, not in skipping it.

---

Sources: Google Search Central documentation on AI Overviews; OpenAI ChatGPT search documentation; Perplexity citation guidelines; Search Engine Land and Ahrefs industry reports, 2025–2026.