Key takeaways
- AI perception is what ChatGPT, Gemini, Perplexity and Google AI Overviews say about your brand when a buyer asks. It is measurable, and it differs by engine.
- Social sources matter very differently per model. Promptwatch Data for Oct 11 to Nov 9, 2026 shows social platforms make up 13.16% of Google AI Overviews citations but only 0.64% of ChatGPT's.
- Pages that state your prices, features and positioning last the longest in ChatGPT citations. News fades fastest, with half of cited news pages gone after 36 days.
- Perception is a claims problem, not just a mentions problem. Track what AI says (pricing, features, support), mark what is wrong, and fix the page it learned it from.
- Most reputation tools were built for humans reading the web. AI visibility platforms are built for models reading it.
Why AI perception became a real reputation channel
For years, brand reputation meant reviews, press coverage and what showed up on page one of Google. That still matters. But a growing share of buyers now ask an assistant first, read one synthesized answer, and never see the ten blue links.
The adoption numbers explain why this is no longer a side project. The Stanford HAI 2026 AI Index reports that organizational AI adoption reached 88% and that 4 in 5 university students now use generative AI. Your future customers are in that group.

There is also a trust problem sitting underneath. BPI's 2026 Reputation Resilience Index found that 81% agree AI makes it too easy to spread false rumors about companies, and only 15% feel truly capable of detecting AI-generated content. Put those together and you get the real issue: a model can repeat a wrong claim about your company with total confidence, and a reader has no easy way to check it.
I think this is the part most marketing teams underrate. A bad review is visible. A bad sentence inside a ChatGPT answer is invisible unless you go looking.
What AI perception actually means
It helps to split perception into four things you can measure separately:
- Whether you are mentioned at all, and how often across a set of buyer prompts.
- Where you land in the answer. First recommendation and fifth recommendation are not the same thing.
- What tone the model uses, and which specific claims it makes ("affordable for small teams", "weak support").
- Which sources it leans on to say those things.
The last point is the one that turns reporting into action. If you know the source, you know what to change. The metrics definitions in the Promptwatch Academy are a decent reference if you want shared vocabulary for visibility, share of voice and sentiment before you start reporting to leadership.
A useful rule of thumb: low share of voice is a coverage problem, late position is a prominence problem, mentions without citations to your own domain are a content or crawlability problem, and weak sentiment is a reputation problem. Each one has a different fix.
What the 2026 data says about where AI gets its opinions
Social sources vary wildly by model
The biggest lesson from tracking across engines is that there is no single "AI" to optimize for. Promptwatch Data's social media citations report (Oct 11 to Nov 9, 2026, published Nov 10) shows how different the engines are:
| Engine | Social platforms' share of citations | Top social source |
|---|---|---|
| Google AI Overviews | 13.16% (October: 12.22%) | YouTube at 6.71% |
| Perplexity | 1.94% (October: 2.71%) | Reddit at 0.82%, LinkedIn 0.66% |
| ChatGPT | 0.64% (October: 0.46%) | Reddit at 0.40% |
AI Overviews cites social media about 20 times as often as ChatGPT. YouTube alone, at 6.71%, is bigger than the other six tracked platforms combined (6.45%). Within AI Overviews, Facebook sits at 2.34%, Reddit at 2.10%, Instagram at 1.02%, LinkedIn at 0.78% and X at 0.02%.
The practical read: if your reputation problem shows up in Google's AI answers, look at YouTube first. If it shows up in Perplexity, Reddit and LinkedIn. If it shows up in ChatGPT, social posts are probably not the cause or the cure. Your own site, listicles, guides and comparisons carry more weight there. And X can stay out of your AI search plan entirely.
This also fits an earlier finding. Promptwatch's data on Reddit citations dropping in ChatGPT showed reddit.com's share of ChatGPT citations fall from roughly 4% to 0.5% on August 14, 2026. It has since crept back to 0.40% in the November report, but it is nowhere near the old level. Anyone who built a ChatGPT reputation plan around Reddit threads in early 2026 had to rethink it.
How long your pages stay cited
Perception is not static. The ChatGPT content half-life report tracked every page cited on July 6, 2026 through October 4. The numbers:
| Content type | Still cited after 30 days | Still cited after 60 days |
|---|---|---|
| Landing pages | 78% | 69% |
| Docs | 76% | 67% |
| Comparisons | 72% | 63% |
| Listicles | n/a | 60% |
| Product pages, reviews, how-tos | n/a | about 55% (half-life 74 to 77 days) |
| News | falls fastest | 41% (half gone by day 36) |
Almost a quarter of news pages were never cited again after day zero. For landing pages, docs, comparisons and listicles, the 50% mark only appears in the last week of the window, so their true half-life is probably near or beyond 90 days.
What this means for reputation: the pages that define how you are described, your pricing, features and positioning pages, are your long-term citation assets. If they are stale or wrong, the error persists for months. Press releases get you a short burst and then disappear from answers.
Authority is not what you think
Another surprise from Promptwatch's August 2026 domain rank data: sites with domain rank 46 to 75 earned almost half of ChatGPT citations, while DR 91 to 100 fell to about 3%. You do not need to be a giant to be a primary source in an answer. A focused, accurate mid-authority site can shape the narrative.
What enterprise data says about the business context
Brands are not tracking this out of curiosity. Deloitte's State of AI in the Enterprise research found that 28% of surveyed leaders now report revenue gains from AI, up from 20% in January, and that 81% say AI has improved the speed of decision-making. When internal teams and customers both lean on assistants to decide faster, the quality of what those assistants say about you affects shortlists before your sales team is involved.

Deloitte also notes that 40% allow AI to execute decisions or actions without human approval to at least a moderate degree. As agents start comparing vendors on a buyer's behalf, a wrong claim about your product stops being an annoyance and starts being a lost deal nobody tells you about.
What brands are learning from tracking
Across teams doing this seriously, a handful of lessons keep repeating.
1. Mentions are the easy metric and the least useful one
Being mentioned feels like progress. But an answer that names you fifth, with a lukewarm description and no link to your site, is a weak result. Track position, sentiment and citation to your own domain alongside mentions.
2. Claims matter more than scores
A single sentiment number hides the cause. The more useful view breaks answers into individual claims, groups the repeated ones, and files them under themes like pricing, features, support and reputation. Then you can see that, for instance, three models say your mobile app lacks features and none of them say so about a competitor.
Once you can see the claim, you can check it. Is it true? If not, which page of yours (or someone else's) says it or implies it? That is what Promptwatch's Brand Narrative and Fact Check are built for: Brand Narrative pulls claims out of tracked responses, and Fact Check lets you mark one inaccurate and then finds the crawled pages on your site that state it or its opposite. Fact Check is on Business and above.
3. Each engine needs its own read
The social citation split above is the clearest example, but it applies to sentiment too. A brand can look great in Perplexity and mediocre in AI Overviews because the sources differ. Aggregating everything into one score hides that. Look at the breakdown by model, country and persona before you decide what to fix.
4. Check that AI can actually read you
Sometimes the narrative is wrong simply because the models cannot reach your best pages. Crawler logs showing which AI bots hit which pages, and where they fail, separate an access problem from a content problem. Free options like the Agentic Readiness Check and the AI crawler URL inspector are a reasonable first pass.
5. Traditional reputation tools only cover half of it
Social listening and review tools tell you what people say. They do not tell you what a model says after reading it. Both matter, but the second view is new for most teams.
How to track your AI reputation: a practical setup
- Write the prompts your buyers actually use. Include category searches ("best X for small teams"), comparison prompts ("X vs Y"), and branded prompts ("is X reliable"). Add personas if your buyers differ by role.
- Pick your markets and languages. Monitors fixed to a country and language keep data from mixing. A brand can be well regarded in the US and absent in Germany.
- Baseline everything. Record visibility, position, sentiment and the top claims per engine before you change anything.
- Trace the sources. For each weak result, list the pages, videos, threads and listicles the model cited.
- Fix in order of half-life. Update long-lived pages first (pricing, docs, comparisons). Then create or refresh comparison pages and listicles. Use news for bursts.
- Re-measure weekly. Models change. ChatGPT's citation behavior shifted notably in 2026, including the site: operator change on August 8 that moved ~0.4% to ~17% of fanout queries overnight.
Tools for tracking AI perception
The market is crowded. Some tools only count mentions, others also explain and fix. Here is how the main categories compare.
| Tool type | Examples | What it answers | Limitation |
|---|---|---|---|
| End-to-end AI visibility platform | Promptwatch | What AI says, why, and how to fix it, then whether it worked | Fact Check and the Content Agent need Business |
| Prompt trackers | Otterly.AI, Peec AI | Was my brand mentioned? | Mostly monitoring only |
| Enterprise visibility suites | Profound, Scrunch | Detailed monitoring for large brands | Higher price points, less execution |
| SEO suite add-ons | Semrush, Ahrefs Brand Radar | AI tracking next to classic SEO | Shallower AI-specific data |
| Social listening and media monitoring | Brand24, Meltwater, Mention | What humans say online | Does not show what models say |
Promptwatch is the one I would start with if you want the whole loop. It collects answers from the real interfaces of ChatGPT, Google AI Overviews, Gemini, Perplexity, Copilot and others, extracts the claims, traces the citations, logs AI crawler visits with IP verification, and then helps you publish fixes and measure whether they stuck. Pricing runs from Essential at $95/month to Business at $579/month, with a 7-day free trial; confirm current numbers at promptwatch.com/pricing. The free AI visibility report is a no-login way to see a first snapshot.

If you are comparing it against other options, the head-to-heads with Profound and Peec are the most requested.

For the human side of reputation, social listening still earns its place.
If you want to browse more options, the GEO software directory at bestgeosoftware.com is a good place to compare them side by side.
Common mistakes
- Treating one engine's result as the whole picture.
- Chasing Reddit and X for ChatGPT visibility when the data shows social is a sliver of its citations.
- Publishing a burst of news and assuming the effect will last. Half of cited news pages are gone within 36 days.
- Letting pricing and feature pages go stale. They are your longest-lived citation assets.
- Reporting a single sentiment score without looking at the individual claims behind it.
- Skipping crawler checks, then rewriting content that models were never able to read.
What to do this quarter
Pick 25 to 50 buyer prompts and baseline them across engines. Read the top five claims for each theme. Pull the sources behind any inaccurate claim and fix the page you control. Update your pricing, docs and comparison pages so the long-lived citations say the right thing. Then add a weekly check. If you want a structured plan, the 90-day GEO program in the Promptwatch Academy lays out the sequence.
If you need hands-on help instead, 1001 SEO Media runs AI search and GEO work for brands and uses Promptwatch for the measurement side.
The short version: AI perception is now something you can look at, break into claims, and change. The brands doing it are not guessing which engine to care about. They read the data per model, fix the pages with the longest shelf life first, and check the result the next week.



