Scope and boundaryThese practices improve eligibility and usefulness. None guarantees indexing, ranking, mention or citation.
5practice groups
3evidence gates per change
0secret markup shortcuts
What the evidence supports
- Accurate visible content comes before structured data or an agent reference file.
- Original data is valuable only when the method, date and limitations are available to inspect.
- Stable questions and raw answers make progress auditable; screenshots without context do not.
Best-practice evidence gates
Use each row as a release check, not a writing prompt.
| Practice | Required evidence | Common failure |
|---|---|---|
| Technical access | Successful fetch, rendered content, canonical and internal discovery | Assuming robots allow means indexed |
| Entity clarity | Consistent name, category, offer, people and profiles | Hidden schema contradicts visible copy |
| Answer usefulness | Direct answer plus decision support and limitations | Generic summary written only for keywords |
| Source quality | Primary sources, first-party method and relevant corroboration | Unsourced superlatives or circular citations |
| Measurement | Exact question, provider, timestamp, answer, citations and sample count | Comparing different prompt sets as a trend |
Apply the practices without bloating the site
01
Consolidate first
Expand the page that already owns the intent before creating a new URL.
02
Add evidence, not filler
Prefer a table, test, dataset or clear limitation to repetitive explanatory copy.
03
Retain raw observations
Make every summary metric traceable to the underlying record.
Methodology
- Selected practices with a direct acceptance test and a current primary source or first-party product basis.
- Removed tactics that depended on hidden content, unsupported schema or guaranteed selection.
- Grouped practices by technical, entity, answer, source and measurement layers.
Limitations
- Provider behavior can change without public documentation.
- Different engines and interfaces may observe different sources.
- Best practices still require market-specific research and editorial judgment.
Questions
Is llms.txt an AI search best practice?
It can help agents find a curated reference, but it is optional and does not replace crawlable pages, sitemaps, accurate content or standard access controls.
How often should AI visibility be measured?
Use a cadence that supports a stable sample and meaningful implementation changes. More frequent noisy runs are not automatically more useful.
Sources and data
- Google Search Essentials: Official search technical requirements, spam policies and core practices.
- OpenAI crawler and user-agent documentation: Official distinction among OpenAI crawler and user-agent controls.
- RadiusRank AI SEO tool: Current public product scope, three-day diagnostic and $79 month-to-month subscription, verified August 14, 2026.
Found a changed price, product limit or source? Email team@radiusrank.com with the official URL. We preserve the snapshot date instead of silently rewriting historical data.