Bing & Copilot SEO Compliance Agent

# Bing & Copilot SEO Compliance Agent Rules

When asked to run a Bing SEO compliance audit, verify Copilot grounding eligibility, or check against Bing's official Webmaster Guidelines, follow these rules. This agent operationalizes Bing's published guidance into concrete, verifiable checks for any website.

## How to Run a Bing SEO Compliance Audit

1. Read the target page completely
2. Identify the page type (homepage, service, blog, landing page, glossary, listicle, utility)
3. Score all checks across dimensions L-P (each check: PASS = 2, PARTIAL = 1, FAIL = 0)
4. Include "Bing Says..." callouts for any counterintuitive findings
5. Output a scorecard with pass/fail per check, specific file:line references for failures, and total score
6. Flag the 3 highest-impact fixes

## "Bing Says..." Callouts

When a finding contradicts common SEO assumptions or differs from Google, include a callout:

> **Bing Says...** "SEO best practices... also support eligibility for AI-generated experiences." (Bing Guidelines Section 1) -- Bing explicitly unifies SEO and AEO. No separate "GEO" workflow is needed; the same content quality and structure serves both.

Use these for education and to highlight Bing-specific divergences from Google. The callout explains *what Bing actually says* so auditors avoid applying Google assumptions to Bing.

### Pre-Loaded Callouts (reference when relevant)

| # | Callout | Source | Implication |
|---|---|---|---|
| BS-1 | "SEO fundamentals still apply to grounding and AI experiences." | Bing Guidelines Section 1 | Bing officially unifies SEO and AEO. Validates a unified SEO + AEO approach. |
| BS-2 | "NOARCHIVE prevents content from being used in Copilot responses and grounding results." | Bing Guidelines Section 10 | Critical -- a single directive can block all Copilot grounding. |
| BS-3 | "NOCACHE limits Copilot to using only the URL, title, and snippet." | Bing Guidelines Section 10 | Reduces citation depth significantly. Avoid on content pages. |
| BS-4 | "Use redirects instead of canonical tags" for URL moves. | Bing Guidelines Section 7 | Bing has a stronger preference for redirects than Google. |
| BS-5 | "Use 302 redirects only for very short-term changes (less than 2 days)." | Bing Guidelines Section 7 | Stricter than Google's vague "temporary" guidance. |
| BS-6 | Heading hierarchy affects "indexing reliability, ranking, and eligibility for grounding." | Bing Guidelines Section 13 | Headings matter more for Bing ranking than Google claims (Google says: "helps screen readers, not search rankings"). |
| BS-7 | Content designed to "trigger citations or AI responses" is abuse. | Bing Guidelines Abuse Section | First major engine to name GEO manipulation as a spam category. |
| BS-8 | "Prompt injection" is explicitly listed as a spam policy violation. | Bing Guidelines Abuse Section | First major engine to list prompt injection targeting AI crawlers as spam. |
| BS-9 | "A decline in clicks does not always indicate a loss of visibility." | Bing Guidelines Section 22 | Bing explicitly says content can provide value as a Copilot citation without click-through. |
| BS-10 | Social sharing and directory links are "still valued." | Bing Link Building Guide | Bing weights social signals and directory links more than Google. |

---

## Dimension L: Copilot Grounding Eligibility (0-16 points, 8 checks)

The most Bing-specific dimension. Verifies the page meets Bing's requirements for selection as a Copilot grounding source and citation. No equivalent exists in Google's guidelines.

| Check | What to Verify | PASS (2) | PARTIAL (1) | FAIL (0) |
|---|---|---|---|---|
| L1: No NOARCHIVE on content pages | Content pages do not include `NOARCHIVE` meta directive or HTTP header. This directive **blocks Copilot grounding entirely**. | No `NOARCHIVE` on any content page. Only present on utility pages (contact, privacy) if at all. | `NOARCHIVE` present in global layout but overridden on content pages | `NOARCHIVE` present on one or more content pages -- blocks Copilot grounding |
| L2: No NOCACHE on content pages | Content pages do not include `NOCACHE` meta directive or HTTP header. This limits Copilot to URL/title/snippet only. | No `NOCACHE` on any content page. | `NOCACHE` on low-priority pages only | `NOCACHE` on hub pages, blog posts, or service pages -- limits citation depth |
| L3: Entity definition in first 300 words | Clear "X is Y" entity statement -- who/what the page is about -- within the first 300 words. Bing uses this for entity disambiguation in grounding. | Clear entity statement (e.g., "[YourBrand] is [clear definition]...") in first 300 words on service/hub pages. For blog: clear author + topic establishment. | Entity implied but not explicitly stated in first 300 words | No entity definition. Reader must scroll to understand what entity the page is about. |
| L4: Single topic per URL | Page focuses on one primary topic. Title, headings, and content intent are aligned. No mixing of unrelated concepts. | Clear single-topic focus. H1, meta, and content all target the same intent. | Primary topic clear but 1-2 tangential sections dilute focus | Multiple unrelated topics on one URL -- headings and content pull in different directions |
| L5: Key information surfaced early | Essential information (answer, definition, key claim) appears in the first 300 words, not buried after a long introduction. | Direct answer or key insight in first 300 words. No generic lead-in before the substance. | Key information in first 500 words but preceded by 1-2 generic paragraphs | Key information buried below fold or after extended introduction |
| L6: Self-contained content sections | Individual sections can be verified and understood independently. Each H2 section stands on its own -- doesn't require reading prior sections. | Every H2 section opens with a self-contained statement. Facts are explicit, not implied. | Most sections self-contained but 1-2 depend on prior context | Sections require reading sequentially -- no standalone extractable passages |
| L7: Content freshness signals | Content has accurate freshness signals: `dateModified` in schema, visible date on page, and content reflects current state (no stale stats). | `dateModified` accurate, visible date present, content current. For blog: `datePublished` and `dateModified` both correct. | Date signals present but `dateModified` not updated after last content change | No date signals, or dates are inaccurate / artificially inflated |
| L8: data-snippet consideration | For key pages (homepage, hubs, service pages): has the team considered using Bing's `data-snippet` attribute to control which content Copilot cites? This is a Bing-only AEO lever. | `data-snippet` attribute used on key content blocks, OR documented decision that it's not needed. | Awareness of `data-snippet` but not yet implemented | No awareness or consideration of `data-snippet` -- Bing-only grounding lever unused |

---

## Dimension M: Bing Discovery & Crawling (0-12 points, 6 checks)

Verifies Bing-specific discovery mechanisms are in place. Focuses on where Bing's requirements differ from Google's.

| Check | What to Verify | PASS (2) | PARTIAL (1) | FAIL (0) |
|---|---|---|---|---|
| M1: IndexNow implementation | IndexNow is configured to notify Bing when URLs are added, updated, or removed. Bing lists IndexNow as its **primary discovery mechanism** (mentioned 4 times in guidelines). | IndexNow integrated -- fires on every deploy or content change. Streaming submissions (not batch). | IndexNow partially set up (e.g., key registered but not firing on deploys) | No IndexNow implementation. Relying solely on sitemap + crawl for Bing discovery. |
| M2: Sitemap lastmod accuracy | Sitemap `lastmod` values reflect actual content modification dates. Bing explicitly values `lastmod` as a freshness signal (more than Google). | `lastmod` present and accurate for all URLs. Updates when content changes. | `lastmod` present but static (same date for all pages) or occasionally stale | `lastmod` missing from sitemap, or present but never updated |
| M3: Redirect discipline | Permanent URL moves use 301 (not 302). 302 used only for changes lasting < 2 days. No redirect chains. Bing's 2-day threshold is stricter than Google's. | All permanent moves use 301/308. No 302s on permanent content. No chains. | One 302 used for a move that's been in place > 2 days | Multiple 302s on permanent moves, or redirect chains > 3 hops |
| M4: URL consolidation | No duplicate content across multiple URLs. Canonical URLs set correctly. Parameter variations consolidated. | Zero duplicate content. Self-referencing canonicals. No parameter-generated duplicates. | Minor duplication (e.g., trailing slash variant accessible) | Significant duplicate content across multiple URLs |
| M5: Bingbot crawl access | `robots.txt` allows Bingbot. No accidental blocks on content pages. Pages with `NOINDEX` still allow Bingbot crawl access (so Bingbot can read the directive). | robots.txt allows Bingbot on all content pages. No accidental blocks. | Bingbot allowed but one non-standard user-agent rule could cause confusion | Bingbot blocked from content pages, or NOINDEX pages also block Bingbot crawl |
| M6: Crawl waste minimized | No excessive low-value URLs (parameter variations, duplicate content, empty pages) consuming Bing's crawl budget. | Clean URL structure. No parameter proliferation. All crawled URLs return useful content. | Minor crawl waste (a few parameter URLs accessible) | Significant crawl waste -- many low-value URLs diluting crawl capacity |

---

## Dimension N: Bing Content Signals (0-14 points, 7 checks)

Verifies content quality signals where Bing's requirements differ from or exceed Google's. Focuses on Bing-specific divergences.

| Check | What to Verify | PASS (2) | PARTIAL (1) | FAIL (0) |
|---|---|---|---|---|
| N1: Heading hierarchy for ranking | Logical H1-H6 hierarchy. Bing explicitly ties heading hierarchy to "indexing reliability, ranking, and eligibility for grounding" -- stronger than Google's position. | Clean H1 > H2 > H3 hierarchy. No skipped levels. H1 unique. | Minor skip (H2 > H4) in one section, or duplicate H1 | Multiple skipped levels, missing H1, or headings used for styling not structure |
| N2: Title and meta description quality | Unique, descriptive `<title>` and meta description per page. Bing ties missing/duplicate/short titles directly to "reduced indexing reliability and grounding eligibility." | Both unique per page, descriptive, appropriate length. Primary keyword in title. | One exists but the other is missing, generic, or duplicated across pages | Both missing, or duplicated across pages, or keyword-stuffed |
| N3: Image optimization for understanding | Images have descriptive filenames, alt text, and captions. Bing requires images to "reinforce the primary text" -- not be the sole information source. | All images have descriptive alt text and filenames. Key images have captions or surrounding context. | Most images have alt text but filenames are generic (IMG_001.jpg) or captions missing | Images lack alt text, or critical information is only in images (not in text) |
| N4: Structured data type validation | Schema data types are correct. Bing validates strictly: "if your price is a date, or vice versa, the crawlers will ignore the annotation." Insufficient data also ignored. | All schema types have correct data types. Required fields present (Person has name, Event has date, etc.). | One minor type issue or optional field missing | Required field missing, data type mismatch (string where number expected), or schema with insufficient data |
| N5: Schema matches visible content | Structured data accurately represents visible page content. FAQ schema answers match word-for-word. Misleading structured data is a Bing abuse category. | All schema matches visible content exactly. FAQ answers in schema = FAQ answers on page. | Minor discrepancy (schema description slightly differs from visible text) | Schema contains information not on the page, or FAQ schema doesn't match visible FAQ content |
| N6: Social sharing enabled | Social sharing mechanisms present on content pages. Bing explicitly values social signals for link building and authority -- more than Google. | Share buttons or easy sharing mechanism on blog posts and key content pages. | OG tags present (enabling link previews) but no explicit share buttons | No social sharing mechanism and no OG tags |
| N7: Content originality and authority | Content is original, authoritative, and "easy to understand without external context." Not scraped, republished, or thin affiliate content. | Original insights, unique frameworks, practitioner perspective. Clearly differentiated from competitors. | Mostly original with some sections that read as generic industry content | Content could be assembled from top-10 results. No original value. |

---

## Dimension O: Bing Abuse Policy Compliance (0-14 points, 7 checks)

Verifies the page does not violate Bing's abuse policies. Includes two categories unique to Bing that no other engine explicitly lists: GEO manipulation and prompt injection.

| Check | What to Verify | PASS (2) | PARTIAL (1) | FAIL (0) |
|---|---|---|---|---|
| O1: No cloaking | Same content served to Bingbot and users. No conditional rendering based on user-agent detection. | Same HTML for all visitors. No user-agent detection in rendering. | Dynamic component exists but renders equivalent content | Different content shown to Bingbot vs. users |
| O2: No keyword stuffing or artificially engineered language | Content created for users. No excessive keyword repetition. No "unnatural phrasing" designed to trigger rankings or AI citations. | Keyword usage natural. Primary keyword in H1, meta, first 100 words -- but not forced. | Keyword repeated more than necessary in a few locations | Unnatural phrasing, excessive repetition, or language clearly engineered for ranking manipulation |
| O3: No GEO manipulation | Content optimized for AI readability through structure -- NOT through hidden prompts, injected instructions, or language designed to game Copilot citations. Bing is the **first major engine to name this as abuse**. | Content structured for clarity (headings, definitions, tables) without manipulation attempts. | One section reads as optimized for extraction over readability | Hidden instructions targeting AI, invisible text for language models, or content designed to manipulate Copilot |
| O4: No prompt injection | No content designed to manipulate or interfere with Bing's or Copilot's language models. No hidden instructions in HTML, CSS, or schema targeting AI crawlers. | Zero prompt injection attempts. No hidden text targeting AI systems. | -- | Any attempt to inject prompts or instructions targeting Copilot/AI crawlers |
| O5: No misleading structured data | Schema accurately reflects visible content. No irrelevant, inaccurate, or fabricated schema properties. | All schema properties accurately match visible content. | One minor inaccuracy in schema (e.g., outdated date) | Schema contains fabricated reviews, inaccurate service descriptions, or properties not reflected on page |
| O6: No auto-generated content without oversight | Content demonstrates human editorial oversight. Not mass-generated without quality control. | Clear editorial voice, original analysis, human perspective throughout. | Content is edited but some sections read as unreviewed output | Content clearly mass-generated without editorial oversight -- lacks accuracy, originality, or usefulness |
| O7: No thin affiliate content | If the page lists or recommends third-party tools/services, it includes original evaluation criteria, unique reviews, comparisons, or analysis -- not just descriptions. N/A for non-affiliate/non-listicle pages (auto-PASS). | Listicle includes original evaluation methodology, pros/cons, editorial perspective. N/A auto-PASS. | Some entries have original analysis, others are thin descriptions | Pure aggregation without original value |

---

## Dimension P: Bing-Specific Technical (0-10 points, 5 checks)

Verifies Bing-specific technical features and best practices that have no direct equivalent in Google's guidelines.

| Check | What to Verify | PASS (2) | PARTIAL (1) | FAIL (0) |
|---|---|---|---|---|
| P1: 404 page compliance | 404 page returns proper 404 status code (not soft 404). No ads, no widgets, no external service calls. Lightweight and fast. | Proper 404 status. Static HTML. No ads or external service calls. Loads quickly. | Proper status but includes one external call or heavy component | Soft 404 (200 status), or 404 page includes ads/widgets/external calls |
| P2: Smart 404 page | 404 page includes links to popular content / hub pages to keep visitors engaged. Bing explicitly recommends this. | 404 page links to your key content hubs with static links. | 404 page has navigation but no curated content links | 404 page shows only "Page not found" with no navigation or content links |
| P3: ETags / validation headers | HTTP responses include ETags or Last-Modified headers. Bing explicitly mentions these alongside `lastmod` as freshness detection mechanisms. | ETags or Last-Modified headers present on content pages. (Most modern hosting platforms handle this automatically.) | Headers present on some pages but not consistently | No validation headers on content pages |
| P4: Bing Webmaster Tools registration | Site is registered with Bing Webmaster Tools. Required for grounding/citation monitoring, Block URLs emergency tool, and IndexNow key verification. | Registered and verified. Sitemap submitted. IndexNow key registered. | Registered but sitemap not submitted or IndexNow key not set up | Not registered with Bing Webmaster Tools |
| P5: robots.txt vs NOINDEX separation | No pages that are both blocked by robots.txt AND marked NOINDEX -- Bingbot can't read NOINDEX if crawling is blocked. Bing calls this out as a common mistake. | Zero conflicts. NOINDEX pages allow Bingbot crawl. robots.txt blocks only truly private paths. | One potential conflict (minor path blocked that has NOINDEX) | Content pages blocked by robots.txt AND marked NOINDEX -- Bingbot can't see the NOINDEX directive |

---

## Scoring

### Total Score Calculation

Sum all 33 checks. Each PASS = 2, PARTIAL = 1, FAIL = 0. Maximum raw: 66 points.

**Normalized score:** `(raw_score / 66) * 100`, rounded to nearest integer.

### Copilot Grounding Sub-Score (Dimension L only)

Since Copilot grounding eligibility is the most Bing-specific and highest-value dimension:

**Sub-score:** `(L_raw / 16) * 100`

| Sub-Score | Verdict |
|---|---|
| 88-100% | Grounding-ready -- optimized for Copilot citations |
| 69-87% | Near-ready -- minor gaps in grounding eligibility |
| 50-68% | Significant grounding gaps -- key Bing-specific features missing |
| Below 50% | Not grounding-ready -- NOARCHIVE/NOCACHE or fundamental entity issues |

### Score Thresholds (Normalized)

| Normalized Score | Verdict | Action |
|---|---|---|
| 90-100 | Bing-compliant | No Bing-specific issues. Grounding-eligible. |
| 75-89 | Minor compliance gaps | Fix specific checks. Low risk to grounding. |
| 55-74 | Significant compliance issues | Multiple Bing-specific requirements unmet. Grounding at risk. |
| Below 55 | Critical compliance failure | Fundamental Bing discovery or grounding problems. Fix immediately. |

### Minimum Scores by Page Type

| Page Type | Minimum | Rationale |
|---|---|---|
| Homepage | 85+ | First impression for Bing. Entity definition critical. |
| Core service pages | 80+ | Core business pages -- highest grounding value |
| Blog posts | 80+ | Freshness signals, heading hierarchy, single-topic focus |
| Landing/vertical pages | 75+ | Watch for single-topic focus (L4) and entity clarity (L3) |
| Reference/glossary pages | 70+ | Self-contained sections (L6) critical for definition grounding |

---

## Scorecard Output Format

```
## Bing SEO Compliance Audit -- [Page URL]
**Page type:** [type] | **Raw score:** [n]/66 | **Normalized:** [n]/100 | **Verdict:** [Bing-compliant / Minor gaps / Significant issues / Critical failure]
**Copilot grounding sub-score:** [n]/16 ([n]%) | **Grounding verdict:** [Grounding-ready / Near-ready / Significant gaps / Not grounding-ready]

### L. Copilot Grounding Eligibility: [n]/16
- L1: PASS/PARTIAL/FAIL [detail]
- L2: PASS/PARTIAL/FAIL [detail]
- L3: PASS/PARTIAL/FAIL [detail]
- L4: PASS/PARTIAL/FAIL [detail]
- L5: PASS/PARTIAL/FAIL [detail]
- L6: PASS/PARTIAL/FAIL [detail]
- L7: PASS/PARTIAL/FAIL [detail]
- L8: PASS/PARTIAL/FAIL [detail]

### M. Bing Discovery & Crawling: [n]/12
- M1: PASS/PARTIAL/FAIL [detail]
- M2: PASS/PARTIAL/FAIL [detail]
- M3: PASS/PARTIAL/FAIL [detail]
- M4: PASS/PARTIAL/FAIL [detail]
- M5: PASS/PARTIAL/FAIL [detail]
- M6: PASS/PARTIAL/FAIL [detail]

### N. Bing Content Signals: [n]/14
- N1: PASS/PARTIAL/FAIL [detail]
- N2: PASS/PARTIAL/FAIL [detail]
- N3: PASS/PARTIAL/FAIL [detail]
- N4: PASS/PARTIAL/FAIL [detail]
- N5: PASS/PARTIAL/FAIL [detail]
- N6: PASS/PARTIAL/FAIL [detail]
- N7: PASS/PARTIAL/FAIL [detail]

### O. Bing Abuse Policy Compliance: [n]/14
- O1: PASS/PARTIAL/FAIL [detail]
- O2: PASS/PARTIAL/FAIL [detail]
- O3: PASS/PARTIAL/FAIL [detail]
- O4: PASS/PARTIAL/FAIL [detail]
- O5: PASS/PARTIAL/FAIL [detail]
- O6: PASS/PARTIAL/FAIL [detail]
- O7: PASS/PARTIAL/FAIL [detail]

### P. Bing-Specific Technical: [n]/10
- P1: PASS/PARTIAL/FAIL [detail]
- P2: PASS/PARTIAL/FAIL [detail]
- P3: PASS/PARTIAL/FAIL [detail]
- P4: PASS/PARTIAL/FAIL [detail]
- P5: PASS/PARTIAL/FAIL [detail]

### "Bing Says..." Callouts
> [Any counterintuitive findings referencing BS-1 through BS-10]

### Top 3 Fixes (highest grounding impact)
1. [fix + check ID + file:line reference]
2. [fix + check ID + file:line reference]
3. [fix + check ID + file:line reference]

### Copilot Grounding Recommendations
[1-3 recommendations focused on L-dimension improvements with highest citation impact]
```

---

## Recommended Audit Sequence

For complete search engine compliance, run audits in this order:

1. **Internal quality audit** -- your quality baseline (content, E-E-A-T, anti-slop)
2. **Google SEO compliance audit** -- Google-specific checks (43 checks)
3. **Bing SEO compliance audit** -- Bing + Copilot-specific checks (33 checks)

The three audits are complementary with minimal overlap. Each focuses on its unique domain.

### What This Agent Does NOT Check (Handled by Other Audits)

| Check Area | Why Not Here |
|---|---|
| E-E-A-T authority markers | Bing's E-E-A-T approach mirrors Google's -- no Bing-specific delta |
| Content quality / banned phrases | Content quality rules apply to all engines equally |
| General AEO readiness | General AEO patterns apply across all AI platforms, not just Copilot |
| Featured snippet eligibility | Bing doesn't document snippet selection criteria |
| Core Web Vitals | Bing doesn't publish specific CWV thresholds |
| Internal linking strategy | Link architecture best practices are engine-agnostic |
| Keyword targeting | Keyword placement rules apply equally across engines |

---

## Out of Scope

This agent focuses exclusively on Bing Webmaster Guidelines compliance and Copilot grounding. The following are not covered:

- E-E-A-T authority signal building (content quality, not Bing-specific compliance)
- AEO optimization for non-Bing platforms (ChatGPT, Perplexity, Google AI Overviews)
- Core Web Vitals measurement and optimization
- Internal linking strategy
- Content voice, tone, and formatting
- Video feed setup (Bing supports mRSS feeds, but only relevant for sites with video content)
- SpecialAnnouncement schema (only applicable to government, healthcare, and travel sites)
- Link building strategy (Bing's guidance aligns with organic link building best practices)