Last updated: 2026-11-17 · Author: Datavinity
To measure AI visibility, track rates across a fixed set of prompts, run repeatedly, over time, not a citation count. The six AI visibility metrics we recommend are brand mention rate, page citation rate, question coverage, top-3 rate, mention consistency and recommendation-prompt share, reported per AI platform. On top of those, add AI share of voice against the competitors you track, then connect it all to AI referral traffic and revenue.
A citation count tells you how often an AI used your page as a source. It doesn’t tell you whether the AI named your brand, recommended you, or did so consistently. AI answers also change from run to run, so a single answer or screenshot proves very little.
Below you’ll find the key numbers, the most common measurement myths, a side-by-side comparison of the metrics, and a ready-to-copy tracking template.
The Verdict: Citation counts alone don’t measure AI visibility. Repeated, per-platform rates for mentions, citations, question coverage and recommendations, tied to traffic and revenue, do.
By the Numbers
| Figure | What it means | Source |
|---|---|---|
| 61.7% | Share of all brand appearances in AI answers that were citations without a brand mention | Semrush (which sells AI visibility tools) and Kevin Indig, Ghost Citations Study, 9 Jun 2026 |
| 87% vs 20.7% | ChatGPT cited a brand’s site in 87% of its appearances but named the brand in only 20.7% | Same study, 9 Jun 2026 |
| Under 1 in 100 | Chance that two responses to the same prompt returned the same list of brands | SparkToro and Gumshoe (Gumshoe sells AI visibility tracking), 27 Jan 2026 |
| About 97 runs | Needed for one prompt on one model to estimate a mention rate within ±10 percentage points (95% confidence) | Gumshoe, which sells AI visibility tracking, 11 Feb 2026 |
AI visibility measurement: myths vs facts
Myth #1: Citation count equals AI visibility
False. A citation and a brand mention are two different outcomes. In the Ghost Citations Study, 61.7% of brand appearances were citations where the brand was never named in the answer. Buyers can read an answer built on your page without ever seeing your name.
Myth #2: One AI answer shows where you stand
False. In research by SparkToro and Gumshoe (Gumshoe sells AI visibility tracking), two responses to the same prompt gave the same list of brands under 1 time in 100, and the same list in the same order closer to 1 time in 1,000. A screenshot is one random draw.
Myth #3: You can track your ranking position in ChatGPT
Misleading. Order changes almost every run, so a single “we rank #2” is noise. What you can track is a top-3 rate: the percentage of answers, across many runs, in which you’re one of the first three brands named.
Myth #4: Visibility on one AI platform carries over to the others
False. Gemini named brands in 83.7% of appearances but cited them in only 21.4%; ChatGPT did the reverse. Engines disagreed on whether to name the brand in 22% of the prompt-and-domain combinations tested. Report each platform separately.
Myth #5: A few runs per prompt give you precise numbers
Misleading. With 30 prompts × 4 AIs × 3 runs a month, your pooled mention rate is accurate to about ±5 points, but a single prompt’s number is only accurate to about ±28 points. Per-prompt results are directional.
Myth #6: AI visibility can be cleanly attributed to revenue
False. You can measure AI referral sessions and conversions and ask customers how they found you, but clean, answer-level attribution doesn’t exist yet. Treat visibility as a leading indicator alongside traffic and pipeline.
Myth #7: Buying ChatGPT Ads improves your AI visibility
False. Ads shown below an AI answer aren’t AI visibility. OpenAI states that advertising doesn’t influence ChatGPT’s answers. Our ChatGPT Ads guide covers the paid side.
AI visibility metrics compared
| Metric | What it measures | How to collect it | Limitations |
|---|---|---|---|
| Brand mention rate | Answers that name your brand ÷ all answers | Run a fixed prompt set; log “brand named?” (text mentions only) | A source link alone doesn’t count; per-prompt numbers are directional |
| Page citation rate | Answers that cite at least one URL on your domain ÷ all answers | Log cited URLs in the run log; Bing AI Performance for Microsoft surfaces | Many citations never name the brand; Bing covers only the surfaces Microsoft lists |
| Question coverage | Tracked prompts where you’re named in at least half of runs ÷ all tracked prompts | Calculated from the run log | The threshold is a choice: write it down and keep it |
| Top-3 rate | Answers where you’re among the first three brands named ÷ all answers | Log your position among brands named | Only meaningful as a percentage across many runs; there’s no stable ranking position |
| Mention consistency | Change in mention rate across runs and across months | Compare “always named” vs “sometimes named” prompts and the monthly trend | Only call it a change if it moves more than the margin of error |
| Recommendation-prompt share | Mention rate on recommendation-type prompts, reported separately | Tag every prompt as Recommendation, Informational or Branded | Needs enough recommendation prompts in the set (we suggest at least a third) |
| AI share of voice | Your brand mentions ÷ all brand mentions (yours plus tracked competitors) in the same answers | Log competitors named in each answer | Tools calculate it differently; don’t compare percentages from two tools |
| AI referral traffic | Clicks from AI assistants to your site | GA4 custom channel group “AI assistants”; utm_source=chatgpt.com | Small and undercounted: many people read an AI answer and never click |
| Revenue and pipeline | Conversions from AI referral traffic | GA4 conversions; CRM “How did you hear about us?” field | Clean, answer-level attribution doesn’t exist yet |
5 Key Takeaways
- Count brand mentions and page citations separately, per platform.
- Measure rates across a fixed prompt set run repeatedly, never a single answer.
- Report a top-3 rate instead of a “ranking position” in AI.
- Respect the margin of error: pooled numbers are solid, per-prompt numbers are directional.
- Connect visibility to AI referral traffic, revenue and self-reported attribution, without promising precise attribution.
AI mentions vs citations: why citation counts mislead
Most “AI visibility” dashboards start with citations, because they’re easy to count. We explained the difference between citations and mentions in our LLM-first content strategy guide; this section is about why you have to count them separately.
Most citations never name the brand
The Ghost Citations study by Semrush and Kevin Indig (built on data from Semrush’s AI Visibility Toolkit) logged 3,981 domain appearances for 115 prompts across 14 countries. It covered ChatGPT, Google AI Overviews, Gemini and Google AI Mode (Semrush, 9 Jun 2026). Each appearance fell into one of three buckets:
| Outcome | Share of all brand appearances |
|---|---|
| Cited as a source, brand not named (“ghost citation”) | 61.7% |
| Cited and named | 13.2% |
| Named, not cited | 25.1% |
Read the denominator carefully. The 61.7% figure is a share of all appearances. It is not a share of citations, even though the study’s headline says “62% of AI citations”.
Only 74.9% of appearances included a citation (61.7% + 13.2%). So among appearances that did include a citation, about 82% (61.7 ÷ 74.9) didn’t name the brand. That 82% is our calculation from Semrush’s published percentages.
Either way, the sample is 115 prompts. Treat the study as a strong signal, not a universal constant.
Each platform behaves differently
In the same study, Gemini and ChatGPT showed opposite patterns (see Myth #4), and engines disagreed on whether to name the brand in 100 of 454 prompt-and-domain combinations tested (22%).
Semrush’s larger AI Visibility Index (126 million US prompts from January to April 2026) reports that ChatGPT cites an average of 15 sources per response, against 3 for Gemini. It also found that on Gemini, the overlap between mentioned brands and cited domains “can be as low as 30%” (Semrush press release, 26 Jun 2026).
Because each assistant draws on different sources, it’s worth deciding which ones your buyers actually use before you build the tracking set. Our breakdown of how ChatGPT, Gemini and Perplexity compare can help you choose.
What kind of question it is matters too
Comparative queries (“best”, “vs”, “recommend”) had a 43.3% brand-mention rate. Informational queries (“what is”, “how does”) had 18%, even though they earned an 89.3% citation rate (Semrush, 9 Jun 2026). That’s why we track recommendation-type prompts separately.
The practical conclusion: a rising citation count can sit alongside a brand that buyers never see named. You need both numbers, measured separately, per platform.
Why one answer tells you nothing: measure rates, not snapshots
If you’ve ever screenshotted a ChatGPT answer that mentions your brand and sent it to your CEO, this section is for you.
The SparkToro and Gumshoe study
SparkToro and Gumshoe asked about 600 volunteers (co-researcher Patrick O’Donnell works at Gumshoe, an AI visibility tracking company, as the study discloses) to run 12 recommendation prompts through ChatGPT, Claude and Google’s AI, 2,961 times in total. Each prompt ran 60–100 times per tool (SparkToro, 27 Jan 2026). Beyond the list-repetition odds above, they found:
- Real users’ prompts barely overlap. When 142 volunteers wrote their own prompts for the same intent, the average semantic similarity was 0.081.
- Frequency is stable even when the lists aren’t. In 994 responses to those 142 different prompts, the leading headphone brands still appeared 55–77% of the time. The study’s conclusion is that “visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric”. But “any tool that gives a ‘ranking position in AI'” can’t be trusted.
The authors note that this isn’t peer-reviewed research.
What this means for the top-3 rate
Top-3 rate is still useful, but only as a percentage across many runs, e.g. “named in the first three in 40% of answers”.
How many runs you need
Gumshoe, which sells AI visibility tracking, published the maths. For one prompt on one model, you need about 97 runs to estimate a mention rate within ±10 percentage points (95% confidence). For ±5 points you need 385 (Gumshoe, 11 Feb 2026).
The workaround is to pool related prompts. Ten related prompts run 39 times each also gets you to about ±5 points.
Here’s what that means for a realistic in-house setup. These figures are our calculation using Gumshoe’s worst-case formula, margin = 0.98 ÷ √n:
| Your monthly setup | Answers collected | Margin of error, all prompts and platforms pooled | Per platform | Per single prompt |
|---|---|---|---|---|
| 30 prompts × 4 AIs × 3 runs | 360 | about ±5 pts | about ±10 pts (90 answers) | about ±28 pts (12 answers) |
How to read that:
- Your overall mention rate is solid.
- Per-platform numbers are directional.
- Per-prompt numbers are only good for spotting where you’re completely absent.
Don’t report a single prompt moving from 33% to 67% as a win.
For a small team, running real category questions three times per platform each month is a workable floor, as long as you keep the margins above in mind.
How to measure AI visibility: the six metrics, defined
These are the six generative engine optimization KPIs (GEO KPIs, not geographic ones) we recommend reporting. Define each metric once, write the definition into your report, and don’t change it mid-year. All six are calculated per AI platform and per period (usually monthly). Use the same prompt set every time.
1. Brand mention rate
Answers that name your brand ÷ all answers collected.
Count a mention only when your brand or product name appears in the answer text. Decide upfront whether product names (e.g. a model number) count as brand mentions, and stick to it. This is the closest thing to “did a buyer see our name?”
2. Page citation rate (plus which pages)
Answers that cite at least one URL on your domain ÷ all answers.
Also log which URLs were cited. A citation of an old blog post, a support page or a retailer listing tells a different story than a citation of your product or comparison page. Pages that are cited but never bring a named mention are your “ghost citation” list. Our LLM-first content strategy guide covers how to rework them.
3. Question coverage
Tracked prompts where you’re named in at least half of the runs ÷ all tracked prompts.
Mention rate can be inflated by a handful of easy prompts. Coverage tells you how much of the buyer’s question space you actually occupy. The “at least half of runs” threshold is our suggestion.
4. Top-3 rate
Answers where you’re among the first three brands named ÷ all answers.
This is how we make “are we a top recommendation?” measurable without pretending there’s a stable ranking position. Report it alongside mention rate. Being named often but rarely first usually means the AI sees you as an alternative, not the default.
For category context, Semrush’s Index found that in consumer electronics the three most visible brands took 76.9% of category visibility, against 41.4% in finance (Semrush press release, 26 Jun 2026). How hard the top 3 is to crack depends heavily on your category.
5. Mention consistency over time
Track two simple things:
- Within a month: the share of prompts where you’re named in all runs, versus some runs. “Always named” is a position; “sometimes named” is a lottery ticket.
- Across months: your mention rate trend, with the margin of error from the table above.
A brand’s place in AI answers isn’t fixed; it has to be maintained. That’s why consistency deserves its own line in the report.
6. Recommendation-prompt share
Tag every prompt by type:
- Recommendation: “best…”, “top…”, “…vs…”, “which should I buy…”, “alternatives to…”
- Informational: “what is…”, “how does…”
- Branded: prompts that contain your name
Then report mention rate and top-3 rate for recommendation prompts on their own, and track what share of your total mentions come from them.
Recommendation prompts are where buying decisions happen, and they behave differently. In the Semrush and Kevin Indig data, comparative queries produced 2.4× the mention rate of informational ones. If most of your mentions come from branded or informational prompts, your “visibility” isn’t reaching new buyers yet.
Layers on top of the six
The six metrics measure presence. Add a competitive layer, then three layers that measure quality and value.
- AI share of voice. Calculate it per platform. Some tools weight position or estimated impressions, so write down your formula and competitor list. Bing’s Citation Share is a different, citation-based measure: your site’s share of all citations for one grounding query (Bing, 16 Jun 2026).
- Accuracy and framing. Is the description correct, current and on-message? Use the run log’s accuracy column to flag answers that are outdated, wrong or off-message, and note when you’re compared on criteria that don’t reflect how you want to be judged.
- AI referral traffic. Clicks from AI assistants to your site (setup below).
- Revenue and pipeline. Conversions from that traffic. Be honest about the gap: clean, answer-level attribution doesn’t exist yet.
The prompt-monitoring template
Copy these tables into a Google Sheet or Excel workbook. Four tabs are enough.
Tab 1: Prompt set (build once, review quarterly)
| Column | What to enter | Example |
|---|---|---|
| Prompt ID | A stable ID you never reuse | P07 |
| Prompt text | Written the way a buyer would ask, in the buyer’s language | “Which mini PC is best for light gaming under $600?” |
| Topic / product line | Groups prompts for pooled reporting | Mini PCs |
| Prompt type | Recommendation / Informational / Branded | Recommendation |
| Funnel stage | Problem / Comparison / Purchase | Comparison |
| Market and language | Where the buyer is | US · EN |
| Source of the prompt | Where you found it: customer emails, sales calls, GSC queries, Bing grounding queries, community threads | Sales call notes |
| Priority | Revenue-critical prompts get checked more often | High |
| Phrasing variants | 2–3 alternative wordings of the same intent | “best small PC for gaming on a budget” |
Tips for the prompt set:
- Size: 20–50 prompts is manageable for a small team.
- Mix: keep at least a third of them recommendation-type.
- Keep it fixed. Don’t add or swap prompts mid-period; changes start a new baseline. How to find and write the prompts themselves is a topic for a separate guide.
Tab 2: Run log (one row per answer)
| Column | What to enter |
|---|---|
| Date, platform, model/mode | e.g. 2026-11-17 · ChatGPT · logged-out, search on |
| Prompt ID · run number | P07 · run 2 of 3 |
| Brand named? (Y/N) | Text mention only |
| Position among brands named | 1, 2, 3… or blank |
| Our domain cited? (Y/N) · cited URLs | Paste URLs |
| Competitors named | Comma-separated |
| Description accurate? | OK / Outdated / Wrong / Off-message, plus a note |
| Full answer saved | Link to a copy of the full answer. Always keep it; you’ll need it to settle disputes |
Keep conditions fixed. Use the same account state (logged out or a clean profile), the same location or VPN, and the same mode (search on or off). Note any model changes the platforms announce.
Tab 3: Monthly scorecard (one row per platform, plus a pooled row)
| Platform | Answers (n) | Mention rate | Citation rate | Question coverage | Top-3 rate | Always-named prompts | Rec-prompt mention rate | Share of voice | Accuracy issues | Margin of error | vs last month |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ChatGPT | 90 | ±10 pts | |||||||||
| Gemini | 90 | ±10 pts | |||||||||
| Google AI Mode | 90 | ±10 pts | |||||||||
| Perplexity | 90 | ±10 pts | |||||||||
| All (pooled) | 360 | ±5 pts |
Tab 4: Owned data (no prompts needed)
| Source | What it gives you | Notes |
|---|---|---|
| Bing Webmaster Tools → AI Performance | Total citations, cited pages, grounding queries and page-level citations across Microsoft Copilot, Bing AI summaries and “select partner AI integrations”. Since June 2026 it also shows intents, topics, citation share and period comparison, in preview (Bing, 10 Feb 2026; Bing, 16 Jun 2026) | First-party, and free. Microsoft says it measures citations, “not rankings, authority, performance, or importance” (Bing help). Grounding queries are a good source of new prompts. Setting up Bing? See our Microsoft Bing Ads guide for the Microsoft side of the ecosystem |
| GA4 custom channel group “AI assistants” | Sessions, engagement and conversions from AI assistants | Google’s own help page gives an example “AI assistants” regex and names ChatGPT, Gemini, Copilot, Claude and Perplexity as examples, but its sample regex has no Claude pattern, so add one (e.g. .claude.) and test it against your own source list. It says to place the channel above Referrals, and notes that custom channel groups apply retroactively (Google Analytics Help) |
utm_source=chatgpt.com | Clicks from ChatGPT search results | OpenAI’s publisher FAQ said ChatGPT adds this parameter automatically (OpenAI Publishers FAQ, archived copy of 4 Oct 2026; the live URL now shows an unrelated article). OpenAI’s current crawler documentation says sites need to allow OAI-SearchBot to appear in ChatGPT search answers (OpenAI crawlers) |
| CRM “How did you hear about us?” field | Self-reported AI influence that never shows up as a click | Add “ChatGPT / AI assistant” as an option |
Checklist: a measurement routine you can sustain
Here’s a routine a small in-house team can keep up:
- Weekly (30 minutes): run your top 5–10 revenue-critical prompts once per platform. Log any wrong or outdated descriptions and fix the source (your page, a listing, a review profile). Check Bing AI Performance for new cited pages and grounding queries.
- Monthly (half a day): run the full prompt set 3 times per platform under fixed conditions, fill in the scorecard, and compare with last month using the margin of error. Pull GA4 “AI assistants” sessions and conversions.
- Quarterly: review the prompt set. Add new questions from sales, support and grounding queries, and retire prompts nobody asks. Re-tag prompt types. Report the trend, not the month.
- Always: save full answers. Never change a metric definition without restating past months. Report pooled numbers with margins, and treat per-prompt numbers as directional.
Connecting AI visibility to traffic and revenue
Visibility metrics are leading indicators. To justify budget, connect them to outcomes. Be explicit about what you can and can’t prove.
- Report AI referral sessions and conversions from your GA4 channel next to the scorecard.
- Watch branded search and direct traffic for lift after visibility gains. That’s correlation, not proof, so label it that way.
- Use self-reported attribution (the CRM field above) to catch influence that analytics misses.
- Don’t promise precise attribution. Treat any tool that claims to attribute revenue to individual AI answers with healthy scepticism.
Semrush’s survey for the Index found that 81% of organizations that fully integrate SEO and AI visibility reported more traffic or leads from AI platforms, against 36% of those running the two separately (Semrush press release, 26 Jun 2026). That’s a self-reported association, not proof of cause. Still, it matches what we see: measurement works best when it’s part of your SEO reporting, not a separate dashboard. More in our guide to making your brand visible in both search worlds.
Other measurement mistakes to avoid
- Blending platforms into one score too early. Report per platform first, then pooled.
- Moving the goalposts. If you change prompts, conditions or definitions, restate past months or start a new baseline.
- Trusting vendor case studies as benchmarks. Anonymised single-client results, measured with a vendor’s own tool, are single cases. Don’t treat them as benchmarks for your own targets.
Our recommendation
Start with 30 prompts, 4 platforms and 3 runs a month, using the four-tab template above.
Report:
- Six numbers per platform: mention rate, citation rate, question coverage, top-3 rate, consistency and recommendation-prompt mention rate.
- Four things alongside them: AI share of voice against your tracked competitors, a pooled row with its margin of error, AI referral sessions and conversions, and a short list of accuracy issues fixed this month.
Judge the trend after three months, not the first snapshot.
About the author: Datavinity is a global growth agency for brands expanding into international markets, with 10+ years of cross-border experience and 90+ brands supported. We run SEO, GEO and paid media for B2B and cross-border teams, in Chinese and English. More about us · GEO & AEO · Consulting
Frequently asked questions
AI mentions vs citations: why track them separately?
Track them separately because they are different outcomes that often don’t happen together. A citation means the AI used your page as a linked source; a mention means it named your brand in the answer text. In a study by Semrush and Kevin Indig, 61.7% of brand appearances were citations without a mention. Report a mention rate and a citation rate separately, per platform.
Which metrics should I use to measure AI visibility?
Use six core metrics: brand mention rate, page citation rate, question coverage, top-3 rate, mention consistency over time, and visibility on recommendation-type prompts. Report them per platform, add AI share of voice against the competitors you track, then connect them to AI referral traffic and revenue.
How many prompts and runs do I need?
You need more than you’d think: for one prompt on one AI, about 97 runs gives a ±10-point estimate, according to Gumshoe, which sells AI visibility tracking. Pooling helps: 30 prompts × 4 AIs × 3 runs a month gives about ±5 points overall, but per-prompt numbers stay directional.
Can I track my “ranking” in ChatGPT?
Not reliably. SparkToro and Gumshoe (an AI visibility tracking vendor) found AI tools return the same list of brands less than 1 in 100 times, and the same order about 1 in 1,000. Track the percentage of answers where you appear, and the percentage where you’re in the first three.
How do I see AI traffic in Google Analytics 4?
Create a custom channel group with an “AI assistants” channel in GA4. Use a regex for assistant domains, and place it above Referrals, following Google’s own example. OpenAI’s publisher FAQ (archived October 2026) said ChatGPT adds utm_source=chatgpt.com to referral links.
Is there a free tool that shows AI citations of my site?
Yes: Bing Webmaster Tools’ AI Performance report is free. It shows citations, cited pages and grounding queries across Microsoft Copilot, Bing AI summaries and select partner integrations. It only covers the surfaces Microsoft lists, so it isn’t a view of every AI assistant.
Can I attribute revenue to AI visibility?
Only partly. You can measure AI referral sessions and conversions and ask customers how they found you. Clean, answer-level attribution doesn’t exist yet. Treat AI visibility as a leading indicator alongside traffic and pipeline.
Sources and methodology
We built this framework from the six metrics we use with clients, then checked each claim against published research. All sources were opened on 16 Nov 2026. Where we calculated something ourselves (the 82% figure and the margin-of-error table), we say so in the text.
| ID | Source | Date | Used for |
|---|---|---|---|
| S1 | Semrush × Kevin Indig: Why 62% of AI citations don’t lead to brand mentions (Ghost Citations Study). Vendor: Semrush sells AI visibility tools | 9 Jun 2026 | 61.7 / 13.2 / 25.1% split; 74.9% cited; platform differences; 22% disagreement; query-type mention rates; sample (3,981 appearances, 115 prompts, 14 countries) |
| S2 | Semrush: 2026 AI Visibility Index press release. Vendor: Semrush sells AI visibility tools | 26 Jun 2026 | 15 vs 3 sources per response; ≤30% overlap on Gemini; top-3 concentration (76.9%, 41.4%); 81% vs 36% survey |
| S3 | SparkToro: AIs are highly inconsistent when recommending brands or products. Co-researcher works at Gumshoe, an AI visibility tracking vendor; disclosed in the study | 27 Jan 2026 | 2,961 runs; under 1 in 100 same list; about 1 in 1,000 same order; prompt similarity 0.081; 55–77% visibility across 994 responses |
| S4 | Gumshoe: How much data do you need to measure AI visibility with confidence?. Vendor: Gumshoe sells AI visibility tracking | 11 Feb 2026 (page footer says “April 2025”; we use the header date) | Sample-size formula; 97 / 385 runs; pooled prompts |
| S5 | Bing: Introducing AI Performance in Bing Webmaster Tools | 10 Feb 2026 | AI Performance report features |
| S6 | Bing: New AI Visibility Insights: Intents, Topics, Citation Share, Compare | 16 Jun 2026 | June 2026 additions (preview); Citation Share definition |
| S7 | Bing Webmaster Tools Help: AI Performance | No date shown | Covered surfaces; “not rankings, authority…” |
| S8 | Google Analytics Help: Custom channel groups | No date shown | AI assistants channel example (sample regex lacks a Claude pattern); place above Referrals; retroactive |
| S9 | OpenAI Publishers and Developers FAQ (archived; Wayback snapshot) | Snapshot 4 Oct 2026 (live URL now redirects to an unrelated article) | utm_source=chatgpt.com |
| S10 | OpenAI: ChatGPT Ads expands to 26 new countries | 21 Oct 2026 | Ads don’t influence answers |
| S11 | OpenAI: Overview of OpenAI crawlers | No date shown | OAI-SearchBot needed for ChatGPT search answers |
If you’d like help setting up the prompt set or turning the findings into content and PR work, see our GEO & AEO service or talk to us.
