SEO That Works for Both Google and AI Search in 2026

Most sites we audit in 2026 are running two SEO programmes without realising it. One team is still chasing Google rankings with title tags and links. Another person, often in a different meeting, is "doing GEO": rewriting pages for ChatGPT, adding an llms.txt file and testing prompts. The two efforts compete for the same budget and sometimes contradict each other.

They don't need to. Google says plainly that there are no special requirements for its AI features, OpenAI and Perplexity publish exactly which crawlers feed their search products, and Microsoft has written down what helps a page get lifted into a Copilot answer. Read those documents side by side and a short list of habits appears that pays off in classic results and in AI answers at the same time.

This guide sets out those ten habits, grouped into three layers: pages machines can reach, content they can quote, and reasons to quote you rather than someone else. For each one we explain what it does in Google, what it does in AI search, and how to check it on your own site. If you want the full task-by-task list, keep our complete SEO checklist open alongside; this article is about which tasks earn double credit.

Why ten habits earn credit in Google rankings and AI answers at once

The overlap is not a coincidence. Google AI Overviews and AI Mode are built on Google's own index. ChatGPT search, Perplexity and Copilot run retrieval systems that fetch and rank web pages before a language model writes the answer. In all of these, a page has to be found, understood and trusted before anything else happens. Google's documentation for site owners is blunt about it:

“There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”

— Google Search Central, AI features and your website

That does not mean nothing changes. AI systems reward some things more heavily than a ranked list of ten links does: self-contained answers, unambiguous facts about your brand, and corroboration on other websites. Semrush's September 2026 roundup cites its own research that pages cited by AI scored about 33% higher on clarity and summarisation and about 31% higher on E-E-A-T signals than pages that only ranked in Google (Semrush, 10 SEO best practices for Google and AI search). Those are correlations from one vendor's dataset, not ranking factors, but they point the same way as the platform documentation.

The table maps each habit to its main effect in each channel. "High" means the habit is a precondition or a strong lever; "Medium" means it helps but rarely decides the outcome on its own.

#	Habit	Effect in classic Google results	Effect in AI answers (AI Overviews, ChatGPT, Perplexity, Copilot)
1	Crawlable, indexable, snippet-eligible pages	High: no index, no ranking	High: each engine's own crawler must be allowed
2	Fast, stable pages with key text in HTML	Medium: Core Web Vitals are one page-experience input	Medium: text hidden in tabs, images or PDFs is harder to extract
3	Topic-based internal linking	High: discovery and relevance	Medium: helps crawlers find supporting pages for fan-out queries
4	Answer-first sections	Medium: featured snippets, better engagement	High: retrieval systems lift passages, not whole pages
5	Structured data that matches visible text	Medium: rich result eligibility	Medium: clarifies entities and attributes
6	Fact-driven freshness	Medium: matters for time-sensitive queries	High for prices, specs and "best X 2026" prompts
7	Clear entity signals for the brand	Medium: knowledge panel, brand SERP	High: decides whether the model names you correctly
8	Original data and first-hand evidence	High: links and helpfulness	High: unique facts are cited because nobody else has them
9	Third-party mentions and reviews	High via links; medium via reputation	High: most brand mentions in AI answers come from other sites
10	Measurement in both channels	Enables every other decision	Enables every other decision

For a deeper look at whether you need separate strategies at all, see our comparison of AEO vs GEO vs SEO. Short version: the labels differ, the groundwork mostly doesn't.

Layer one: pages that Googlebot, OAI-SearchBot and PerplexityBot can reach and read

Habit 1: Allow the right crawlers and keep pages snippet-eligible

Google's rule for AI Overviews and AI Mode is that a page must be indexed and eligible to be shown in Search with a snippet. That has a practical consequence many sites miss: a nosnippet robots directive, or a very low max-snippet value, limits what Google can show from the page in AI features as well as in normal results. Use those controls deliberately, not as a leftover from an old template.

Outside Google, each AI search product uses its own crawler, and each one is controlled separately in robots.txt. Blocking the training crawler does not have to block the search crawler. This is the setup we check first:

User agent	Operator	What it does (per the operator's docs)	Typical setting for a business site
Googlebot	Google	Crawls for Search, including AI Overviews and AI Mode	Allow
Google-Extended	Google	A control token for training and grounding in some other Google AI systems; not a separate crawler	Business decision; does not affect Search
OAI-SearchBot	OpenAI	Surfaces sites in ChatGPT search features	Allow if you want ChatGPT search visibility
GPTBot	OpenAI	Collects content that may be used to train foundation models	Business decision; independent of OAI-SearchBot
ChatGPT-User	OpenAI	Fetches pages when a user's request needs them; robots.txt may not apply	Do not block at the firewall by accident
PerplexityBot	Perplexity	Surfaces and links sites in Perplexity results; not used for model training	Allow
Perplexity-User	Perplexity	User-initiated fetches; generally ignores robots.txt	Monitor, do not rely on robots.txt

Sources: OpenAI crawler documentation, Perplexity crawler documentation and Google's AI features page. OpenAI notes that robots.txt changes take roughly 24 hours to be reflected in its search results, which makes this one of the fastest fixes in SEO when a block is found.

The frequent failure is not robots.txt at all. It is a CDN or security product returning a challenge page to unfamiliar bots. Google's own guidance asks site owners to allow crawling both in robots.txt and in any CDN or hosting layer in front of the site. Our guide to blocking or allowing AI crawlers walks through the server-log checks.

Quick test, in order:

1. Open /robots.txt and search for each user agent in the table, plus any User-agent: * group with a broad Disallow.
2. In Search Console, run URL Inspection on your top five money pages and confirm "URL is on Google" and that the live test renders your main text.
3. Search the raw HTML (not the rendered DOM) for nosnippet, max-snippet and data-nosnippet.
4. Filter 30 days of server logs for OAI-SearchBot and PerplexityBot. If they request pages and receive 403 or 429 responses, your firewall is the problem.
5. Check that canonical tags point to the version you want cited, especially on sites with parameter or print URLs.

Habit 2: Keep pages fast and keep the important text in HTML

Page experience is a modest ranking input in Google, and Core Web Vitals are measured at the 75th percentile of real visits. The "good" thresholds are an LCP of 2.5 seconds or less, an INP (which replaced FID in March 2024) of 200 milliseconds or less, and a CLS of 0.1 or less (web.dev, How the Core Web Vitals thresholds were defined).

For AI systems the bigger issue is where text lives. Microsoft's October 2025 guidance for inclusion in AI answers warns against hiding key information in tabs or expandable menus, and against core details locked in PDFs or images without HTML text. A pricing table published as a screenshot, a spec sheet that exists only as a PDF, or a comparison that loads after a click all reduce the chance that a retrieval system picks up the fact you most want repeated.

• Render critical text server-side or in the initial HTML; client-side-only content is a risk for crawlers that do little JavaScript rendering.
• Publish an HTML version of any PDF that carries prices, specifications or policies.
• Use real <table> markup for comparisons rather than styled div grids or images.
• Write alt text that states what an image shows, especially for charts with numbers.

Habit 3: Link internally by topic, not by keyword

Internal links help Google discover pages and understand how a site is organised, and Google lists them among its recommendations for AI features. They matter in a second way for AI search: both AI Overviews and AI Mode can run a "query fan-out", issuing several related searches across subtopics and drawing supporting pages from each. A site that covers a topic in a connected cluster (core guide, comparison, pricing, how-to, glossary) gives those sub-searches more of your pages to find.

Practical rule: every page in a cluster should link to the cluster hub with a descriptive anchor, and the hub should link out to every child page with an anchor that says what the reader will learn there. Orphan pages, and hubs that only link to the five newest posts, are the patterns we fix most often.

Layer two: content an answer engine can lift without losing its meaning

Habit 4: Put the answer in the first two sentences of each section

Search engines rank pages; AI systems usually quote passages. A passage that only makes sense after reading the three paragraphs above it is a weak candidate. Microsoft put the reason in one line:

“AI assistants don’t read a page top to bottom like a person would.”

— Krishna Madhavan, Microsoft Bing, Optimizing Your Content for Inclusion in AI Search Answers

The edit that helps both channels is simple. Under each H2 or H3, open with one or two sentences that answer the heading directly, including the subject by name rather than "it" or "this". Then add the detail, caveats and examples. Readers who skim get the answer; Google gets a clean candidate for a featured snippet; an AI system gets a self-contained passage.

Before and after, using a hypothetical accounting firm page:

• Before: "As we mentioned, it depends on several factors, which we'll look at below."
• After: "A UK sole trader must register for Self Assessment if self-employment income is over the trading allowance; the steps below show how to register and what records to keep."

Treat the second version as a pattern, not legal advice: the point is that the sentence names the subject, the condition and the outcome.

Habit 5: Use structured data to label facts readers can already see

Structured data still earns rich results in Google for supported types, and Google's AI features guidance asks that structured data match the visible text. It is not a shortcut to AI citations, and the list of rewarded types keeps shrinking: HowTo rich results went in 2023, and FAQ rich results stopped appearing in Google on 7 May 2026, with the related Search Console reports withdrawn in the following months (Priority Pixels). FAQPage markup left on a page does no harm, but it no longer produces a visual result in Google.

Schema type	Google benefit in October 2026	Why it still helps AI systems	Our recommendation
Organization (with logo, sameAs, contact)	Logo and knowledge panel signals	Ties the brand name to official profiles	Every site, homepage or about page
Product / Offer	Product snippets and merchant listings	Machine-readable price, availability, ratings	All ecommerce product pages
LocalBusiness	Supports local understanding	Address, hours and service area as data	Every location page
Article (author, dates)	Article features, date understanding	Author and freshness attribution	Blog and news content
FAQPage	None visible since May 2026	Other parsers may still read it	Leave existing markup; don't build projects around it
HowTo	None since 2023	Limited	Not a priority

Our schema markup guide has the JSON-LD templates. The rule that matters for both channels: if a value is in the markup, it must be on the page, and it must be current.

Habit 6: Update pages when the facts change, not when the calendar does

Changing a "last updated" date without changing the content is common and pointless. What matters is that prices, product names, statistics, screenshots and year-specific claims are right. AI assistants answering "best payroll software for small business" or "how much does a loft conversion cost" repeat whatever numbers they retrieve, so a stale page doesn't just lose rank; it can put an out-of-date price in front of a buyer under your name. That problem is common enough that we wrote a separate piece on why AI still describes your old brand.

A workable refresh rule set:

• Pages with prices, plans or specifications: review whenever the fact changes, and at least quarterly.
• Pages with a year in the title: review in the first month of the year and when a major platform change lands.
• Evergreen explainers: review annually, or when Search Console shows impressions falling for the page's main queries.
• Only change the visible "updated" date and dateModified when the substance changed.

Layer three: reasons for Google and AI tools to cite you over a competitor

Habit 7: Make your brand an unambiguous entity

Language models are good at text and weak at disambiguation. If your company name is shared with another business, if your about page describes the company three different ways, or if directories still show a pre-rebrand name, the model may merge you with someone else or describe you vaguely. Google faces the same problem when building a knowledge panel.

The fix is consistency in places machines read:

1. Write one canonical description: what you are, who you serve, where, and the one thing that distinguishes you. Use it word for word on the homepage, about page, Organization schema, LinkedIn and main directories.
2. Add sameAs links in Organization markup to your official profiles.
3. Give every product or service its own page with a consistent name. Don't call the same plan "Pro" on the pricing page and "Professional" in the docs.
4. Correct the top five third-party profiles that rank for your brand name.

Branded queries are where entity confusion shows first. Our analysis of AI Overviews on brand searches explains how to audit what Google's AI says when people search your name.

Habit 8: Publish at least one fact nobody else has

When ten pages say the same thing, an AI system has no reason to cite any particular one, and Google has no reason to rank yours first. Original data breaks the tie. Google's people-first content guidance asks the question directly:

“Does the content provide original information, reporting, research, or analysis?”

— Google Search Central, Creating helpful, reliable, people-first content

Original does not have to mean a 5,000-respondent survey. For most of our clients it means one of these: anonymised averages from their own job or order data; a pricing benchmark from quotes they issued; a test they ran (load times, product durability, response times); or a dated photo record of a project. State the method and sample size in a sentence, put the key number in a heading or the first line of a section, and date it. See our guide to original research that earns links for formats.

Habit 9: Earn mentions on the third-party pages assistants actually read

Backlinks have always been a Google ranking input. In AI search, unlinked mentions and reviews matter too, because the model reads what other sites say about you. Semrush's roundup cites AirOps research from 2025 finding that about 85% of brand mentions in AI search came from third-party pages, and its own study finding that 62% of AI citations were "ghost citations" in which a site was cited but the brand was not named (Semrush). Both figures come from vendor research rather than independent audits, but they match what we see when we check prompts by hand: list articles, review platforms, forums and trade media dominate the sources.

The practical consequence is that digital PR, review generation and correcting directory listings now count twice. Our workflow for AI citation outreach covers how to find the pages AI tools cite for your prompts and how to approach their owners. For B2B companies, where review platforms and comparison articles carry even more weight, see our guide to B2B answer engine optimization.

Habit 10: One reporting view across Search Console, Bing and GA4

Measurement is a habit, not an afterthought, because without it you cannot tell whether a change helped Google, AI answers, both or neither. The tools have improved a lot since 2025:

Tool	What it shows	What it can't show
Google Search Console, Performance report	Clicks and impressions, with AI Overviews and AI Mode traffic counted in the "Web" totals	No separate breakdown of AI feature impressions
Bing Webmaster Tools, AI Performance (public preview since February 2026)	Citations in Copilot and Bing AI summaries, cited pages, sample grounding queries	ChatGPT, Perplexity and Google activity
GA4 AI Assistant channel (default channel group since May 2026)	Sessions referred by recognised AI assistants such as ChatGPT, Gemini and Claude	Visits with no referrer, which still land in Direct
AI visibility trackers (Semrush and others)	Mentions, citations and share of voice across a fixed prompt set	True prompt volumes; results are samples
"How did you hear about us?" field	Self-reported AI influence on leads	Precise attribution

Sources: Bing Webmaster Blog, TechWyse on the GA4 AI Assistant channel, Google's AI features page. TechWyse notes that the GA4 channel "does not resolve all AI attribution gaps", so treat it as a floor. Our guide to tracking AI traffic in GA4 explains how to check for sessions that slip into Direct, and the AI visibility dashboard guide shows how to put prompt tracking next to traffic.

A monthly review that fits on one page:

1. Search Console: clicks, impressions and average position for the 20 pages that drive revenue, compared with the same month last year.
2. Branded impressions in Search Console as a proxy for demand created elsewhere, including in AI answers.
3. GA4: AI Assistant sessions, key events and conversion rate by landing page.
4. Bing AI Performance: total citations and which pages are cited.
5. Prompt tracking: mention rate and citation rate on 30 to 100 fixed prompts, with notes on what changed in the answers.

Worked example: applying the ten habits to a 40-page accounting firm site

This is a hypothetical example built to show the arithmetic; the numbers are illustrative, not client results.

A regional accounting firm has 40 indexed pages: 8 service pages, 4 location pages and 28 blog posts. Baseline for the last 90 days: 6,000 organic clicks from Google, 90 sessions in the GA4 AI Assistant channel, and a mention in 3 of 40 tracked prompts (7.5%), such as "accountant for contractors in Leeds" and "how much does a small business accountant cost".

The audit finds five problems, each tied to a habit:

Finding	Habit	Fix	Effort (hours)
Security plugin returns 403 to OAI-SearchBot and PerplexityBot	1	Allow-list the user agents and published IP ranges	1
Fees are only in a downloadable PDF	2, 6	Publish an HTML fees page with a table and "from" prices	6
Service pages open with 200 words of company history	4	Rewrite first paragraph of each section to answer the heading	12
Three different firm descriptions across site and directories	7	One canonical description, Organization schema with sameAs, fix top directories	5
No original data anywhere	8	Publish anonymised average fees and turnaround times from last year's 300 clients	10

Total effort: 34 hours. If, after a quarter, the firm is mentioned in 10 of the 40 prompts, the mention rate rises from 7.5% to 25%. If AI Assistant sessions grow from 90 to 180 per quarter and convert to enquiries at 4% rather than the 2% organic average, that is about 7 enquiries from AI referrals instead of about 4 (90 × 4% = 3.6; 180 × 4% = 7.2). Those conversion rates are assumptions for the example. The point of writing it down is that every input can be measured with the tools in the previous section, so the firm can see which fixes paid off.

Note what is absent from the list: no new "AI pages", no llms.txt file, no rewriting every post for chatbots. Every fix also helps the Google rankings that still drive most of the firm's 6,000 clicks.

Tactics we would drop because they help neither channel

Some popular advice in 2026 costs time without a documented benefit in either Google or AI search:

• Special files for AI. Google says you don't need new machine-readable files or AI text files to appear in its AI features. A file nobody's documentation asks for is not a priority.
• Schema for content that isn't on the page. It contradicts Google's guidance and confuses parsers.
• Mass-produced "answer" pages. Thin pages written only to match prompts do not meet Google's people-first standard and offer AI systems nothing they can't find elsewhere.
• Date-bumping. Changing timestamps without changing content is a freshness signal with nothing behind it.
• Blocking every AI bot by default. If you want visibility in ChatGPT or Perplexity search, blocking their search crawlers removes you from the candidate set. Decide training and search access separately.

Questions SEOs ask about optimising for Google and AI answers at the same time

Do I need a separate GEO or AEO strategy?

Usually not a separate strategy, but a few extra checks: AI crawler access, answer-first formatting, entity consistency and third-party mentions. Those sit inside a normal SEO programme.

Does ranking in Google help me appear in ChatGPT or Perplexity?

Indirectly. They run their own crawlers and retrieval, but the qualities that help a page rank (relevance, clarity, authority, links) also make it a stronger retrieval candidate. Google's AI Overviews and AI Mode draw directly on Google's own index, so the link there is closer.

Should I block GPTBot?

That is a decision about model training, not search. OpenAI documents that GPTBot and OAI-SearchBot are controlled independently, so you can block training and still allow ChatGPT search.

Is FAQ schema still worth adding?

Not for Google: FAQ rich results stopped appearing in May 2026. Existing markup is harmless. Visible question-and-answer sections on the page are still useful to readers and to AI systems.

How quickly do changes show up in AI answers?

Crawler unblocks can register within about a day in ChatGPT search, according to OpenAI. Content and reputation changes take weeks to months, because they depend on recrawling and on other sites updating what they say about you.

Where to start on your site in the next 30 days

Pick the habits in order of cost and certainty. Week one: crawler access, snippet controls and the five-step quick test above. Week two: rewrite the opening of each section on your ten highest-revenue pages so it answers the heading. Week three: write your canonical brand description and fix Organization schema and the main directories. Week four: set up the monthly measurement view and choose one original data point to publish next quarter.

If you would rather have a team run this across both channels, our SEO services cover the technical, content and authority work together, and our content writing team can handle the answer-first rewrites. To scope it for your site, get a free quote or request a custom quote based on your page count and goals.

Further reading on SEO for Google and AI search

• B2B Answer Engine Optimization: Getting Shortlisted by AI
• Google's AI Contribution Pilot: Google Pays Publishers for AI Content
• AI Citation Outreach: A Semrush and Claude Workflow

Sources for this Google and AI search SEO guide

1. Semrush — 10 SEO best practices for Google and AI search (September 2026)
2. Google Search Central — AI features and your website
3. Google Search Central — Creating helpful, reliable, people-first content
4. OpenAI — Overview of OpenAI crawlers
5. Perplexity — Perplexity crawlers
6. Microsoft Advertising — Optimizing your content for inclusion in AI search answers (October 2025)
7. Bing Webmaster Blog — Introducing AI Performance in Bing Webmaster Tools (February 2026)
8. web.dev — How the Core Web Vitals metrics thresholds were defined
9. TechWyse — Google Analytics adds native AI Assistant channel (May 2026)
10. Priority Pixels — Google ends FAQ rich results in Search (May 2026)