AI品牌提及追踪
从手动到自动化的追踪体系

# 追踪关键词体系设计示例

keyword_categories = {
    'brand_keywords': {
        'description': '品牌词 - 测试AI是否认识你的品牌',
        'count': '5-10个',
        'examples': [
            'XX科技有限公司',
            'XX科技',
            'SmartCRM',
            'SmartCRM Pro',
            'XX CRM系统'
        ],
        'query_frequency': '每周1次',
        'success_criteria': 'AI准确描述品牌核心业务和产品'
    },
    'industry_keywords': {
        'description': '行业推荐词 - 测试AI是否推荐你的品牌',
        'count': '10-15个',
        'examples': [
            '最好的企业CRM软件',
            '企业级项目管理工具推荐',
            'SaaS CRM对比',
            '适合中小企业的CRM系统',
            'AI驱动的CRM软件'
        ],
        'query_frequency': '每周2次',
        'success_criteria': '品牌出现在AI推荐列表中'
    },
    'comparison_keywords': {
        'description': '竞品对比词 - 测试AI对比中的表现',
        'count': '5-10个',
        'examples': [
            'SmartCRM vs 竞品A',
            'XX CRM和竞品B哪个好',
            '国产CRM软件对比'
        ],
        'query_frequency': '每周1次',
        'success_criteria': 'AI对比中突出本品优势'
    }
}

for cat, info in keyword_categories.items():
    print(f'{cat}: {info['count']} | 频率: {info['query_frequency']}')
    for ex in info['examples']:
        print(f'  - {ex}')

# Playwright自动化品牌提及追踪脚本
from playwright.sync_api import sync_playwright
import json, time, re

BRAND_NAMES = ['SmartCRM', 'XX科技', 'SmartCRM Pro']
COMPETITORS = ['竞品A', '竞品B', '竞品C']
KEYWORDS = [
    '企业CRM推荐',
    '最好的项目管理软件',
    'SmartCRM怎么样',
    '企业级CRM对比'
]
PLATFORMS = {
    'chatgpt': 'https://chatgpt.com',
    'kimi': 'https://kimi.moonshot.cn',
}

def check_mention(text, names):
    '''检查文本中是否提及品牌或竞品'''
    mentions = {}
    for name in names:
        if name.lower() in text.lower():
            mentions[name] = True
    return mentions

def analyze_sentiment(text, brand_name):
    '''简单的情感分析（实际应用中建议使用LLM API）'''
    positive_words = ['优秀', '推荐', '领先', '创新', '好用', '强大']
    negative_words = ['差', '问题', '缺点', '不足', '投诉', '负面']
    
    # 找到品牌提及附近的文本
    idx = text.lower().find(brand_name.lower())
    if idx == -1:
        return 'not_mentioned'
    context = text[max(0, idx-100):idx+200]
    
    pos_count = sum(1 for w in positive_words if w in context)
    neg_count = sum(1 for w in negative_words if w in context)
    
    if pos_count > neg_count:
        return 'positive'
    elif neg_count > pos_count:
        return 'negative'
    else:
        return 'neutral'

results = []
with sync_playwright() as p:
    browser = p.chromium.launch(headless=True)
    for plat_name, plat_url in PLATFORMS.items():
        for kw in KEYWORDS:
            page = browser.new_page()
            page.goto(plat_url)
            time.sleep(2)
            page.fill('textarea', kw)
            page.press('textarea', 'Enter')
            page.wait_for_timeout(20000)  # 等待AI生成回答
            response_text = page.text_content('body')
            
            brand_mentions = check_mention(response_text, BRAND_NAMES)
            comp_mentions = check_mention(response_text, COMPETITORS)
            
            for brand in BRAND_NAMES:
                if brand_mentions.get(brand):
                    sentiment = analyze_sentiment(response_text, brand)
                else:
                    sentiment = 'not_mentioned'
                
                results.append({
                    'platform': plat_name,
                    'keyword': kw,
                    'brand': brand,
                    'mentioned': brand_mentions.get(brand, False),
                    'sentiment': sentiment,
                    'competitors_mentioned': [c for c in COMPETITORS if comp_mentions.get(c)]
                })
            page.close()
    browser.close()

with open('mention_tracking_results.json', 'w', encoding='utf-8') as f:
    json.dump(results, f, ensure_ascii=False, indent=2)
print(f'追踪完成，共{len(results)}条记录')

# 品牌提及追踪数据分析
import json
from collections import defaultdict

def analyze_tracking_data(results_file):
    with open(results_file, 'r', encoding='utf-8') as f:
        data = json.load(f)
    
    # 按平台统计引用率
    platform_stats = defaultdict(lambda: {'total': 0, 'mentioned': 0})
    for r in data:
        platform_stats[r['platform']]['total'] += 1
        if r['mentioned']:
            platform_stats[r['platform']]['mentioned'] += 1
    
    print('=== 平台引用率对比 ===')
    for plat, stats in platform_stats.items():
        rate = stats['mentioned'] / stats['total'] * 100
        print(f'{plat}: {rate:.1f}% ({stats['mentioned']}/{stats['total']})')
    
    # 按关键词分类统计
    keyword_stats = defaultdict(lambda: {'total': 0, 'mentioned': 0})
    for r in data:
        keyword_stats[r['keyword']]['total'] += 1
        if r['mentioned']:
            keyword_stats[r['keyword']]['mentioned'] += 1
    
    print('\n=== 关键词引用率对比 ===')
    for kw, stats in sorted(keyword_stats.items(), key=lambda x: x[1]['mentioned']/x[1]['total'], reverse=True):
        rate = stats['mentioned'] / stats['total'] * 100
        print(f'{kw}: {rate:.1f}% ({stats['mentioned']}/{stats['total']})')
    
    # 情感分布
    sentiment_stats = defaultdict(int)
    for r in data:
        if r['mentioned']:
            sentiment_stats[r['sentiment']] += 1
    
    print('\n=== 情感分布 ===')
    total_mentioned = sum(sentiment_stats.values())
    for sent, count in sentiment_stats.items():
        pct = count / total_mentioned * 100 if total_mentioned > 0 else 0
        print(f'{sent}: {count} ({pct:.1f}%)')
    
    # 竞品共现分析
    competitor_cooccur = defaultdict(int)
    for r in data:
        if r['mentioned'] and r.get('competitors_mentioned'):
            for comp in r['competitors_mentioned']:
                competitor_cooccur[comp] += 1
    
    print('\n=== 竞品共现频率 ===')
    for comp, count in sorted(competitor_cooccur.items(), key=lambda x: x[1], reverse=True):
        print(f'{comp}: {count}次')

analyze_tracking_data('mention_tracking_results.json')