Optimizing without a baseline
If you never recorded where you started, you cannot tell whether anything worked.
Guide / Workflow
How to rank in AI search: an eight-step workflow you can run this month
The short answerYou cannot rank in AI search the way you rank in a list, and no one can guarantee a citation. You can improve your odds with a repeatable workflow: choose the prompts that matter, record a baseline, confirm crawl and snippet access, rewrite key pages to answer directly, add sourced evidence, mark up structured data, earn independent mentions, and recheck on a schedule.
The workflow
Eight steps in order. Steps 1 and 2 are the ones people skip, and the ones that make the rest measurable.
- 1
Pick 15 to 30 prompts that map to revenue
Write the questions a buyer would type: what is, best for, versus, how to, pricing. Include your category and your competitors' names. Fewer, better prompts beat hundreds of vague ones.
- 2
Record a baseline
Run each prompt on the engines you care about, several times. Classify every answer as absent, mentioned, cited, or recommended. Save the stored answers. This is your before picture.
- 3
Confirm access and eligibility
Make sure target pages are indexable, not blocked by robots rules or noindex, and not restricted from snippets. Google says pages must be indexed and snippet-eligible to be supporting links (source). Check crawler rules for other engines deliberately; OpenAI documents its crawlers and their purposes (source).
- 4
Rewrite the answering page to answer first
For each prompt, pick the one page that should answer it. Put a one- to three-sentence direct answer immediately under the heading. Use descriptive headings, short paragraphs, and a table or list where the content is a comparison or sequence.
- 5
Add evidence you can defend
State facts specifically, name the source, link it, and give the date. Add first-hand detail only you have: your process, your data, your constraints. Remove claims you cannot support.
- 6
Mark up structured data that matches the page
Add Organization, Article, and FAQPage markup only where the page genuinely contains those things. Google's structured-data policies require markup to represent the page's visible content (general guidelines), and markup does not guarantee a search feature.
- 7
Earn independent corroboration
Engines draw on what other sources say about you. Make sure your name, category, and facts are consistent on profiles, directories, and review sites, and pursue relevant coverage and reviews. Do not buy fake ones.
- 8
Recheck on a schedule and keep the diff
Re-sample the same prompts every two to four weeks. Compare to baseline. Keep what moved, revisit what did not, and note that run-to-run variation is normal.
Define success before you start
Decide which outcome you are after for each prompt. A mention is progress for awareness prompts. A citation matters for traffic. A recommendation matters for comparison prompts close to purchase.
- 01
Organic ranking
Where a URL sits in a list of classic search results.
Cause: relevance, quality, links, technical health.
- 02
Brand mention
An AI answer names your brand in its text, with no link.
Cause: what the web broadly says about you.
- 03
Citation
An AI answer attaches your page as a linked source.
Cause: your page answers the question clearly, with evidence.
- 04
Recommendation
An AI answer explicitly suggests you as a choice.
Cause: strong fit plus corroboration from independent sources.
Each outcome can happen without the others. A page can rank first and never be cited; a brand can be recommended from a page that ranks nowhere.
Page checklist before you publish
Use it as a final review on every answering page.
| Check | What good looks like |
|---|---|
| Opening | Direct answer within the first paragraph under the H1 or H2 |
| Headings | Each describes the section, ideally phrased as the question |
| Evidence | Specific facts with named, linked sources and dates |
| Entities | Brand, product, and people named in full and consistently |
| Structure | Tables for comparisons, numbered lists for sequences |
| Schema | Valid JSON-LD that matches visible content |
| Access | Indexable, snippet-eligible, in the sitemap, linked internally |
| Freshness | Visible update date and facts that are still true |
Mistakes that waste months
Patterns that look like progress and are not.
Reading a single run
One answer is an anecdote. Sample repeatedly.
Publishing unsourced statistics
Invented or vague figures hurt trust with readers and any system that checks claims.
Blocking crawlers by accident
A blanket disallow can remove you from a product's results. Check rules with the AI robots checker.
Treating llms.txt as a lever
It is a proposal, not a guarantee. Fine to add, wrong to rely on.
Where RankZoro helps at each step
The workflow works without any tool. These are the places a platform saves time.
Steps 1, 2, 8: queries and citation tracking
AI Citation Tracking stores your query set, samples engine citations, cited URLs and positions, and supports domain comparison. Review brand mentions and explicit recommendations separately.
Step 3: access
Run a site audit and the AI robots checker for crawl and snippet issues.
Steps 4 to 6: page work
The GEO Content Optimizer flags weak openings and missing sources and generates structured data.
Step 7: corroboration
Competitor and source gap views show which independent sites engines lean on in your category.
Sources
External references used on this page.
- Google Search Central: AI features and your websiteEligibility and SEO fundamentals for AI Overviews and AI Mode.
- OpenAI: Overview of OpenAI crawlersDocuments crawler user agents and purposes for robots rules.
- Google Search Central: structured data policiesGeneral technical and quality guidelines for structured data.
- Aggarwal et al.: GEO: Generative Engine Optimization (arXiv)Research on content strategies and generative engine visibility.
Frequently asked questions
How do I get my website cited by ChatGPT or Perplexity?
There is no switch. Make target pages crawlable and indexable, answer the question directly, support claims with named sources, keep entity names consistent, and earn independent mentions. Then sample the relevant prompts over time to see whether results change.
How long does it take to see results?
It varies by engine, site, and competition. Re-check every two to four weeks and compare against your baseline rather than expecting a fixed timeline.
Do I need llms.txt?
It is optional and no engine is obliged to honor it. If you add one, treat it as documentation, not a ranking lever. The free llms.txt generator can create one.
Should I block AI crawlers?
That is a business decision. Blocking a crawler can remove you from that product's results, while allowing it may expose your content to training or retrieval. Review each crawler's documented purpose and decide per bot.
Is schema markup required?
No. Google says no special markup is needed for AI features. Valid, accurate schema still helps machines understand a page, and FAQ markup is only appropriate for content that is actually on the page.
Can I do this without a tool?
Yes, with a spreadsheet and manual prompt checks. A platform mainly saves time on repeated sampling, competitor comparison, and page audits.
What if an AI answer describes my business wrongly?
Find which sources it used, correct the facts on your own pages and major profiles, and recheck. You cannot edit the answer directly, but you can fix the information it draws from.
Run your first sample before you change anything
Start with a citation baseline in RankZoro, then manually review the same neutral questions for brand mentions and explicit recommendations. Build the plan from evidence, not guesses.