How to build your 2027 AI visibility plan
2 November 2026
A framework for brand teams, with a worked example.
Your customers are already asking ChatGPT, Gemini, Claude, Perplexity, Copilot and Google's AI Overviews which brand to choose. Most brand teams going into 2027 budget planning have no line for it, no owner and no measure.
This guide shows how to build a simple AI visibility plan you can take into budget planning. It uses a fictional UK brand, Fernly (plant-based snacks), to show what each step looks like in practice.
Why AI visibility needs its own plan
AI answers don't work like search results. Instead of ten links, people get one answer and a handful of brand names. Those names are recommended, mentioned in passing, named with a caution, or left out.
And AI mostly forms its view of your brand from other people's pages. In Seenford's research across 1,223 AI answers and 8,835 cited sources in 15 sectors (23 Sep to 23 Oct 2026):
- Fewer than 1 in 10 cited sources (7.9%) were brands' own websites.
- Ranked "best of" lists made up 29.3% of cited pages. Product pages made up 3.5%.
- Engines behave very differently. Brand-owned sources were 21.9% of ChatGPT's citations, but 2.1% of Claude's. Forums were 28.6% of Google AI Overviews' sources, but 0.4% of Copilot's.
- Questions about a brand by name carried a caution 14.2% of the time, against 2.1% for "what's the best…" questions.
That makes AI visibility a brand, PR and content job, not only an SEO one. It needs a plan that cuts across teams.
Step 1: Start with a baseline
Every plan starts with what AI says about you today, across all six engines. Measure a few things:
| Measure | What it tells you |
|---|---|
| Named in unprompted answers | Do people asking "what's the best…" hear about you? |
| Recommended when named | Does AI rate you once it knows you're relevant? |
| Answers with a caution | What worries does AI repeat about you? |
| Your own site among cited sources | How much of your story AI takes from you |
| Your category leader's unprompted mentions | The gap you need to close |
The Fernly example: named unprompted by 1 of 6 engines, recommended by 4 of 6 when named, cautions from 2 engines (price and availability), and its own site at 3% of cited sources. The category leader was named unprompted by all six.
The pattern is common: AI speaks well of a brand when asked directly, but rarely puts it forward when someone asks for a recommendation.
Step 2: Set goals you can report every quarter
Pick three measures and set quarterly targets. Keep them modest. AI answers vary from week to week, so report a range over several weekly checks rather than one number. Results also lag effort: work in Q1 usually shows in Q3 and Q4.
Fernly's 2027 goals:
| Measure | Today | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|---|
| Engines naming Fernly unprompted (of 6) | 1 | 1 | 2 | 3 | 4 |
| Recommendation rate when named | 67% | 67% | 70% | 75% | 80% |
| Answers with a caution (of 6) | 2 | 2 | 1 | 1 | 0 |
Step 3: Choose four bets, no more
Each bet needs an owner, a cost, a date and an outcome. Most brands' bets fall into four areas:
- Get onto the pages AI reads. The "best of" lists, review sites and forums engines cite for your category. Usually a PR and partnerships job.
- Be a clear source of truth. Plain, current facts on your own site that AI can quote, so engines stop disagreeing about you. Usually digital and content.
- Answer the caveats. If AI keeps repeating the same concern, such as price or availability, deal with it where AI looks. Usually ecommerce, product or customer teams.
- Measure and report. Weekly tracking across all six engines, with a quarterly review. Usually the brand team.
Which lists matter, which facts are wrong and which caveats come up will be different for every brand and every engine. That's what your baseline is for.
Step 4: Measure what you can't click
AI answers rarely send a trackable click, so last-click reporting will undercount this work. Report proxies instead:
- Visibility: engines naming you unprompted, by priority question.
- Quality: recommendation and caution rates when you're named.
- Influence: your share of the sources AI cites, and which lists you appear on.
- Business signals: branded search and direct traffic, alongside the above.
A one-page quarterly report to leadership is enough: what moved, what didn't, and what you'll change.
Step 5: Make the ask
Keep it short:
- Approve £[total] for 2027 across the four bets.
- Book the people and development time the bets need, early in the year.
- Agree a 15-minute quarterly review with leadership.
Most plans need no new headcount. The work sits with existing teams and agencies. The plan gives them priorities and a measure.
Start with your baseline
Every number in a plan like this starts with one question: what does AI say about your brand today?
Get a free Quick Report, a quick snapshot of your brand across ChatGPT, Gemini, Claude, Perplexity, Copilot and Google AI Overviews.
Want help building the plan? Seenford's audit gives you the baseline, a prioritised fix list and weekly tracking. Book a demo
Research: Seenford analysis of 1,223 AI engine answers and 8,835 cited sources across 15 sectors, 23 Sep to 23 Oct 2026. Fernly is a fictional brand, and its figures are illustrative.