Best Practices
Each article here follows one reader from start to finish: what to check first, what to build, how to measure and what to leave alone. The steps link to the entries that hold the evidence, so you can read the route in one sitting and come back for the details.
Every article marks which steps are documented by the platforms, which rest on industry observation and which are still guesses.
Slides for sharing: a short overview deck for decision makers, and decks by website, role, business situation and market. Each deck has a PDF download. Each slide repeats a claim from these guides with its evidence level and source.
Entries in this topic
Guides
- A 90-day GEO plan for a company brand, starting from zeroSpend the first 90 days building a baseline and fixing foundations, not chasing mentions. Measure where you stand, make sure AI search can reach your site, rework a few key pages, and get listed accurately where third parties compare your category. Expect visible movement after month three.
- Before you hire a GEO agencyA credible GEO provider sells measurable work, such as baselines, content and technical fixes, and repeatable re-tests. It does not sell guaranteed AI rankings. Start with a small paid pilot, write sampling rules, deliverables, data ownership and banned tactics into the contract, and judge results on shares measured over many runs.
- Cross-border brands: splitting GEO between China and abroadThe two markets share almost no AI platforms or sources, so treat them as two programmes on one set of facts. Give each language its own URL, open each market's crawlers, write natively for each, and build third-party presence where each market's assistants read. Measure each market separately.
- For content editors: writing articles AI answers can quoteWrite something only you can write, then make it easy to lift. Answer the question in the first paragraph, use descriptive headings, give numbers with sources, sign and date the piece honestly. Google says there is no need to chop pages into fragments for AI.
- For marketing leads: measuring GEO and reporting it upwardNo single number measures GEO. Combine the platforms' own AI reports, AI referral traffic in analytics, and a fixed prompt panel run enough times to be meaningful. Report trends with ranges and sample sizes, tie them to leads where you can, and never promise rankings or a fixed position in AI answers.
- GEO best practices for a WordPress siteMost WordPress GEO problems are settings, not content. Check the search visibility box, the robots.txt your plugins write, and any firewall or CDN that blocks bots. Then keep one sitemap and one structured data source, and measure from logs and search reports.
- GEO best practices for documentation sites and developer productsAI tools quote docs they can fetch as plain text and match to the right version. Allow search and user-fetch crawlers, serve every page as server-rendered HTML with a Markdown twin, keep one canonical URL per version, and treat llms.txt as a cheap convenience rather than a ranking lever.
- GEO best practices for technical and SEO leadsAI search can only cite pages it can reach, read and quote. Check indexing, crawler tokens, CDN and firewall rules, snippet controls and server rendering, then confirm verified crawler visits in the logs. Turn the checks into automated tests so a deploy or a CDN default change cannot silently undo them.
- GEO for healthcare, legal and financial businessesIn health, money and law, an inaccurate AI answer can hurt people, and an over-promising page can break advertising law. Put compliance review first, publish verifiable facts with named experts and dates, keep content balanced, and keep it current. Promotional content that dodges ad review is the main legal risk.
- GEO for local businessesNobody can guarantee that an AI assistant recommends your shop first. Make your business easy to find and identify. Keep one accurate name and set of details on every map and review platform, earn genuine reviews, look after the pages others publish about you, and measure the share of answers that mention you.
- GEO in China: Doubao, Yuanbao, Qwen, ERNIE, DeepSeek and KimiChinese assistants each lean on their own group's content, and few sources are shared between them. Test in the real apps, publish in each ecosystem your customers use, keep the search engines able to crawl you, and look after third-party pages. Everything here is observed, not documented, so re-test often.
- GEO outside China: Google AI Overviews, ChatGPT, Perplexity and CopilotEach product has its own gate. Google AI features need an indexed, snippet-eligible page; ChatGPT search needs OAI-SearchBot; Perplexity needs PerplexityBot; Copilot draws on Bing's index and live searches. Open all four gates, then build content and third-party coverage, and measure with the two official reports.
- JavaScript sites (Next.js, Vue and others): making sure AI crawlers can read your contentGooglebot renders JavaScript, but a 2024 study found the major AI crawlers did not. Put the main content, links and metadata in the server HTML using SSR, static generation or pre-rendering. Then verify with a no-JavaScript fetch and server logs.
- Minimum viable GEO for one person with a few hours a weekWith little time, do the things platforms actually document. Make sure AI search can reach and quote your pages, fix your few most important pages, then spend the rest on original, useful content and a small monthly measurement routine. Skip special AI files and rewrite schemes.
- New brands: why AI doesn't mention you yet, and what to do firstAI answers favour brands the web already writes about, and a model may know nothing about you. Make your site reachable and unambiguous first, publish one clear facts page, then win narrow questions and your first independent coverage. Measure with enough repeats, because small brands appear unevenly.
- Online stores (Shopify and others): getting products recommended and cited by AI shoppingAI shopping draws on three sources: crawled pages, product feeds and your live site. Keep all three reachable, complete and in agreement on price and stock. Write product pages that answer real buyer questions, and earn genuine reviews instead of inventing them.
- PR and brand teams: how press, encyclopedias, Q&A sites and reviews shape AI answersAI answers lean on what others publish about you, so earned coverage matters. Find which sources answers in your category cite, then earn real coverage, reviews and encyclopedia entries under each platform's rules. Paid placements, undisclosed edits and fake reviews break platform policies and, in China, count as AI data poisoning.
- When AI gets your brand wrong: find it, fix it, prevent itMost wrong answers trace back to a source: an outdated page, a third-party listing, a confused entity or a model's training cutoff. Fix the source first, then use platform feedback channels, which are limited. Prevent repeats with one maintained facts page and regular checks.